Senate convenes hearing on AI, affordability and innovation

[0:00] I thought it was the Democrats not being here yet. [0:02] But anyways, good point. [0:05] Good point is how do we get maybe that was all LBR. [0:11] OK, she's there. [0:12] OK, how do we get artificial intelligence right? [0:17] That is the ultimate question. [0:19] I think we are struggling with as a nation and frankly, as a committee. [0:23] That means making sure AI strengthens our economy, supports small businesses and protects consumers. [0:30] It means keeping America ahead of China and our adversaries. [0:35] It means using American innovation to help families facing the high cost of housing, borrowing and everyday life. [0:43] AI is a powerful tool. [0:46] Like any powerful tool, it brings both opportunity and risk. [0:50] We should not run from this potential, but we should also not ignore the risks. [0:55] We should ask the right questions, listen to the experts before us and make sure America remains ahead. [1:04] For the people we serve in South Carolina and across this nation, the real question is whether this technology makes life better. [1:11] Can it help a family in North Charleston save time and money on a mortgage? [1:16] Can it help a small business in Greenville become more competitive, protect a senior in Columbia from fraud, bring new jobs to communities like Spartanburg? [1:27] If we get this right, the answer can be yes. [1:30] AI will also change the way Americans work. [1:33] That does not mean we should fear the future. [1:37] It means we need to make sure workers have the tools, training and opportunity to benefit from it. [1:44] The future of work should not be something America fears. [1:48] It actually should be something America leads. [1:52] AI solutions that help increase affordability for American families will not come from government bureaucracy. [1:58] They will come from entrepreneurs, innovators, workers and small business owners willing to use these new tools to build, compete and solve problems. [2:08] This is where AI can be part of the solution. [2:12] Roughly 20 percent of small businesses in the United States today are already using AI to help lower the cost of doing business, [2:19] which means prices are going to be lower for consumers, and in the financial services sector, AI can reduce friction and, frankly, increase speed. [2:31] It can improve customer service, help detect fraud faster, help small businesses do more with less. [2:40] AI should be a tool that strengthens American workers, not one that makes them feel replaceable. [2:48] There is dignity in work, and if America leads, AI can help workers build skills, support their families, and create new opportunities in South Carolina and all across this great nation. [3:04] But responsible AI innovation matters. [3:07] And when we consider regulation, we must ask whether it will make it harder for a community bank to lend, a startup to grow, or a family to access credit. [3:17] We need thoughtful and well-informed solutions that protect Americans and mitigate major risks without hampering innovation or pushing it overseas. [3:27] That is why this hearing is so important and, frankly, timely. [3:34] AI is already changing our economy and our markets. [3:37] We must make sure the next chapter of responsible innovation is written here in America, not in China. [3:44] Cybersecurity and national security must be taken seriously. [3:49] The United States cannot afford to let China or any other adversary gain a technological edge in artificial intelligence. [3:58] This includes global markets where China and Huawei are pursuing an aggressive agenda to displace U.S. technology around the world. [4:06] The world's tech stack, including powerful AI tools, must be built on trusted U.S. networks and technology. [4:17] We have to carefully craft export control policy that is clear and concise, allowing U.S. companies to compete in the global market while ensuring we are not handing our adversaries tools they can use against us. [4:32] This hearing builds on the work we have already started through member roundtables. [4:37] Today, we will hear from experts who can help this committee better understand the opportunities, the risks, and the choices before us. [4:46] I look forward to your testimonies. [4:48] If America leads, AI can help lower costs, improve affordability, expand opportunity, strengthen small businesses, protect our financial system, and keep America secure. [5:01] That is the future we should be fighting for. [5:04] Ranking mover? [5:05] The floor is yours. [5:07] Thank you, Mr. Chairman, and thank you for holding this hearing. [5:10] Artificial intelligence has extraordinary potential and presents extraordinary risks. [5:16] Americans across this country are deeply concerned about how artificial intelligence will affect their lives. [5:23] Americans are also concerned about what looks like corruption in plain sight, driving the Trump administration's AI policy. [5:34] Influential tech giants are concentrating ownership over powerful technologies and attempting to rewrite the rules that govern them. [5:43] A corrupt Trump administration basically is giving them whatever they want. [5:48] The impact on American families is massive. [5:52] First, AI companies are on a spending spree to build data centers. [5:58] This is the infrastructure that supports AI, and families in areas surrounding these data centers are already paying a price. [6:07] Residential electricity prices, which have risen more than 16% since Donald Trump took office, are expected to climb as much as 40% by 2030. [6:20] Nearly one in three households is struggling to pay their utility bills. [6:26] The Trump administration's response has been a weak request for voluntary commitments from big corporations, and that leaves families footing the bill. [6:38] Second, AI companies are addicted to debt. [6:42] Early in AI development, these companies financed their own investments, but now they are borrowing upwards of a trillion dollars to finance spending on data centers, chips, and other AI infrastructure. [6:58] This means that if AI tumbles, it will bring down many in private credit, and that in turn will risk bringing down giant banks. [7:16] The possible failure of AI companies looms larger every day. [7:22] AI companies that are unable to increase revenues and service their massive debt loads could trigger massive losses in the financial sector and another 2008-style crash. [7:35] That, in turn, could wipe out huge amounts of value from Americans' retirement accounts. [7:43] Third, Americans are worried about what will happen to their jobs in a world of rapidly improving AI capabilities. [7:51] AI has already started reshaping the workforce. [7:55] While we don't know what the long-term effects on unemployment will be, we cannot wait around to see how many jobs will be eliminated before we start to prepare for the coming upheaval. [8:09] We have a decision to make. [8:12] Either AI will profit a handful of the richest people, making them even richer, or the benefits of AI will be used for all Americans. [8:24] If we continue to do nothing, the billionaires will suck up all the value from AI. [8:31] But if we start now to tax AI and to use that money to fund free education, apprenticeships, and encourage new job creation for the workforce, [8:42] we can share the benefits of AI across our country. [8:47] Before we hear from our witnesses, I want to raise one other issue. [8:51] China is also working feverishly to develop AI, including AI weapons systems that can be turned against the United States. [9:02] One company, NVIDIA, makes the chips that are powering AI's advances. [9:10] In recent years, NVIDIA's valuation has ballooned to nearly $5 trillion. [9:15] That makes NVIDIA, the world's biggest company, and NVIDIA's CEO, Jensen Huang, a very powerful person. [9:26] And Mr. Huang wants to sell NVIDIA's chips to China, which will profit his own company and advance China's AI development. [9:35] Now, I invited NVIDIA's CEO, Jensen Huang, to testify at this hearing, and he refused to join us. [9:44] I invited Mr. Huang, because he and NVIDIA are lobbying hard, [9:50] to get a blank check from the Trump administration to allow the sale of advanced chips to China. [9:56] Mr. Huang's efforts have alarmed senior military leaders here in the United States [10:02] who have warned that such technology could strengthen China's military and put our own national security at risk. [10:13] No matter. [10:14] In Trump's D.C., money talks, and President Trump appears to be giving policy handouts to the highest bidder. [10:23] If it's not plain, old-fashioned bribery, it sure smells like it. [10:29] Mr. Huang and NVIDIA have been active here in D.C. [10:33] NVIDIA donated a million dollars to the Trump inauguration. [10:36] Mr. Huang made a big donation to Trump's favorite project, his gold-encrusted ballroom. [10:43] Mr. Huang wheedled his way into a seat on Air Force One on the president's summit with President Xi in China. [10:51] Mr. Huang even managed to score a meeting with President Xi himself, along with President Trump, [10:58] about NVIDIA's business interests. [11:01] All that sucking up to the president has paid off big time for Mr. Huang. [11:06] America had strong export controls that would have blocked NVIDIA from selling these chips to China, [11:14] but the Trump administration is relaxing these rules. [11:18] The Trump administration has also turned a blind eye as massive numbers of these chips have been sold elsewhere, [11:26] then somehow magically made it to China. [11:30] The American people deserve to hear Mr. Huang answer hard questions from both sides [11:38] about the role he has played in undermining our national security. [11:43] AI holds tremendous promise, but Americans are rightly concerned that it could further rig our economy. [11:54] Congress can no longer be a bystander as the risks from AI grow nearer [11:59] and millions more American chips are sent to China. [12:04] It is time for Congress to stand up so that the American people can share in the success of this technology [12:11] and not just be left holding the bag for whatever goes wrong. [12:15] Thank you, Mr. Chairman. [12:18] I'll now introduce the witnesses quickly. [12:21] Today we have before us Mike Flynn, Senior Vice President of ITI. [12:25] David Feith, Senior Fellow at the Hudson Institute. [12:31] We have Will Reinhart, who is a Senior Fellow at the American Enterprise Institute. [12:36] We also have Dr. Sarah Myers West, who is the Co-Executive Director of AI Now. [12:44] First, we'll hear from Mr. Flynn. [12:46] You now have the floor for five minutes. [12:49] Chairman Scott, Ranking Member Warren, Distinguished Members of the Senate Banking Committee. [12:56] My name is Mike Flynn, and I'm a Senior Vice President at the Information Technology Industry Council, or ITI. [13:03] Thank you for inviting me to testify today. [13:05] ITI represents 80 companies at the center of American innovation, including companies that operate at every layer of the AI stack, [13:15] chip makers and designers, leading frontier AI labs and cloud service providers, cybersecurity firms, [13:23] and the software developers that bring AI to consumers in the U.S. and around the world. [13:29] ITI's members are investing hundreds of billions of dollars in the United States to build semiconductor fabs, [13:36] conduct research and development for new AI applications, and drive innovation at every layer of the AI stack. [13:45] Artificial intelligence is the most significant driver of economic growth in a generation, [13:51] and the competition within ITI's membership and the broader IT ecosystem is fierce. [13:57] Frontier AI models rely on thousands of highly skilled engineers, advanced semiconductors, [14:03] with features many times smaller than the width of a human hair, and networking that operates at the speed of light. [14:10] To name only a few of the vital inputs needed to build and run a frontier AI model, [14:16] we should view the American AI stack as an asset to U.S. economic security. [14:22] Congress must not take this for granted. [14:24] U.S. companies need clarity and certainty to sign multimillion-dollar contracts with customers, [14:31] spend billions of dollars on custom chips, [14:34] and continue the trillion-dollar investments in America that are making the U.S. the undisputed AI leader. [14:41] The question before this committee is how to protect American AI leadership [14:46] while promoting adoption of the U.S. AI stack around the world. [14:51] The AI supply chain is highly specialized, resource-intensive, and globally complex. [14:59] An AI server in an American data center providing AI to customers in South Carolina [15:05] depends on memory from South Korea, chips from Arizona made with tools designed in Massachusetts. [15:12] A semiconductor may touch as many as 70 countries before it reaches a consumer. [15:17] Key parts of the supply chain are based in allied nations in Europe and Asia, [15:22] especially South Korea, Taiwan, and Japan. [15:26] Because of this reality, the United States must work with like-minded nations [15:31] on economic security policies, like export controls. [15:35] The agency Congress has tasked with this important work is the Bureau of Industry and Security, or BIS. [15:41] To meet the demand of this moment, BIS needs appropriate resources and oversight. [15:46] Recently, ITI endorsed six bills that advance this through the House Foreign Affairs Committee. [15:54] These bills are a down payment to improve BIS's ability to fight diversion of controlled items, [16:01] ensure timely licensing reviews, and develop AI export controls side-by-side with industry experts. [16:08] Effective export control policy and enforcement must be part of a broader strategy [16:13] that enables the U.S. technology sector to remain agile and competitive in global markets for the long term. [16:21] With the remainder of my time, I'd like to touch on one other issue that I know is top of mind for this committee, [16:27] and that is the cybersecurity of financial institutions. [16:30] Frontier AI models are developing and accelerating the pace [16:34] and presenting new challenges for cybersecurity professionals. [16:37] That is why it is more important than ever that industry and government work together [16:42] to address these emerging risks. [16:44] One thing Congress can do right now is to ensure a long-term reauthorization [16:50] of the Cybersecurity Information Sharing Act, or CISA-15. [16:54] This is a critical law that protects the ability of organizations [16:58] to share cyber threat information and vulnerability information, [17:02] not only with the government, but with one another. [17:04] There is more work to be done, and we welcome the partnership of this committee [17:10] on these serious questions of U.S. economic security. [17:13] Thank you, Mr. Chairman, for convening this hearing, and I look forward to the committee's questions. [17:17] Thank you, sir. [17:18] Mr. Feith. [17:21] Chairman Scott, Ranking Member Warren, and committee members, [17:24] thank you for the opportunity to testify. [17:27] My focus is AI competition with China [17:30] and how export controls can help preserve American security and other interests. [17:34] We face a historic opportunity today with three aspects. [17:38] First, the AI industrial revolution will shape not just American civilization, but human history. [17:45] It's hard to overstate the need to get AI policy right, [17:48] especially regarding our strategic challenge from China. [17:51] Second, we're at a hinge point in Trump administration AI policymaking, [17:56] which rightly recognizes that America must win the AI race with China. [18:00] Recent events involving the super-powerful mythos model, [18:04] ship smuggling, and export control loopholes [18:06] show why we need better strategy, rules, enforcement, and oversight. [18:11] Third, this committee has a particularly key role to play. [18:15] Members here and across Congress have introduced an admirable set of bills [18:19] to strengthen America's export control system, bolster national security, [18:23] and assert Congress's role in this key area of policy. [18:25] What's at stake in the strategic contest over AI? [18:30] One essential feature is the personalization of technological power. [18:35] This is in some ways very positive, in some ways very threatening. [18:39] AI offers personalized medicine, but also personalized warfare, [18:45] and especially ominously personalized state power. [18:48] Will this technology ultimately advantage systems that protect human freedom [18:52] or systems that stifle and tyrannize? [18:55] There is some good news to start with. [18:59] In the U.S.-China competition in AI, [19:01] the decisive factor is likely to be computing power, or compute. [19:05] Here, the United States and our allies enjoy a substantial advantage. [19:09] Today, China's best AI chips are much less capable than America's, [19:13] and thanks largely to export controls, their chips are falling further behind. [19:18] Compute is the lifeblood of AI, [19:20] and China has less of it at lower quality than do the United States and our allies. [19:24] That makes compute China's principal choke point in the AI contest. [19:30] But this advantage is not self-sustaining. [19:32] It will last only if U.S. and allied policy preserves it. [19:36] This is where export controls come in. [19:38] Unfortunately, current U.S. export control policy has not been equal to the task. [19:43] The Trump administration's top-line policy [19:45] is to prevent America's most advanced chips from going to China. [19:49] President Trump has said this repeatedly. [19:51] His administration's AI action plan also calls for tougher export controls [19:54] and more allied cooperation. [19:56] But these policies are not yet realized. [19:59] Instead, our stand on exporting advanced AI chips to China has become less clear. [20:05] Commerce Department entity listings haven't kept pace. [20:08] Controls on semiconductor manufacturing equipment, [20:10] perhaps the most important issue of all, remain loose. [20:14] Remote access remains unaddressed. [20:17] Amid signs of massive smuggling, especially via Southeast Asia, [20:19] we have seen a sensational federal indictment but little adjustment from our export control system. [20:25] And we have only recently learned that alarming loopholes in U.S. rules were evidently created last year, [20:30] allowing sales of the most advanced U.S. chips to Chinese firms outside of China [20:34] and maybe also allowing China access to leading-edge fabrication at allied firms like TSMC. [20:40] Finally, on distillation attacks, supercharged IP theft for the AI age, [20:47] we risk repeating 20 years of past failure to counter China's hacking, [20:53] hence the historic opportunity before this committee. [20:57] I'd recommend enacting the set of complementary bills now before the Senate, [21:00] including the AI Overwatch Act on exports to China, [21:03] the Match Act on equipment, the Remote Access Security Act, [21:07] the CHIP Security Act on location verification, [21:09] the Deterring AI Model Theft Act on distillation, [21:13] and the CHIP-Equip Act, protecting American fabs at home. [21:17] Each addresses a key piece of the puzzle. [21:20] Also, enhance oversight of commerce. [21:24] Draw on the classified record and real evidence from export control enforcement in the field. [21:30] Finally, beyond export controls, [21:32] you have an opportunity to codify the Commerce Department's ICTS authority [21:36] and help protect U.S. data center infrastructure from sabotage by Chinese-exposed hardware. [21:44] Key technological developments have changed world history. [21:48] Iron, gunpowder, nuclear weapons. [21:51] AI is another. [21:53] The future now turns on whether America, [21:56] Western civilization, and liberal democracy will lead the age of AI, [21:59] or communism, oppression, and state power will. [22:03] That is why these issues matter so much. [22:05] America should innovate at home, export confidently to allies and trusted partners, [22:10] and deny China the capabilities it needs to overtake us. [22:14] Thank you. [22:15] I look forward to your questions. [22:17] Thanks, sir. [22:17] Mr. Reinhart, the floor is yours. [22:19] Chairman, ranking member, committee members. [22:23] As a kid, I was obsessed with my grandfather's red tractor, a classic Farmall H. [22:28] For decades, you could find it parked in between the garden and the barn, [22:31] basically, you know, always pointed towards the house, [22:34] acting like a sphinx, protecting the cornfields. [22:36] I've since learned that the Farmall was truly revolutionary for the farms and families [22:40] where I grew up in central Illinois. [22:44] Tractors first spread through the wheat belt in the 1920s, [22:47] but it took another two decades for the tech to be adopted for use in the Corn Belt. [22:52] Farmalls were especially important for small to medium-sized family farms [22:56] because they were high-quality but affordable tractors [22:58] that could do a range of different tasks. [23:00] It took years, however, for those farms to truly adapt to the new technology [23:04] and to make it useful and to change how farming was done in the Midwest. [23:10] As I see it, AI is in its wheat belt stage. [23:13] It is having an impact in some very narrow and important categories, [23:16] but yet it is yet to be transformative to the broader economy. [23:20] As I've detailed in my written testimony, [23:23] one of the biggest trends that I'm seeing currently right now in AI [23:26] is that it's a skill equalizer. [23:28] In customer support, software development, legal analysis, writing, and education, [23:33] AI continues to do its best by boosting those who are at the bottom. [23:37] But when tasks fall outside the model's current capabilities, [23:40] when workers are already at the frontier, [23:42] AI can do more to hurt performance than to help it. [23:47] Importantly, judgment still matters. [23:50] So as I see it, Congress really should be thinking about two things, [23:53] scenario planning and better data sources. [23:56] For this committee in particular, planning around agentic AI could be very useful. [24:00] Agents are only now getting deployed, and I know there's been some discussion of this, [24:03] but they could have some incredible and important impacts in finance and banking and commerce. [24:08] That's something that I have been focused a lot on, and I believe you should as well. [24:12] But we also need better data sources. [24:14] Right now, the data on AI usage only has effectively a 15% response rate. [24:19] There are better ways to collect information, especially through payroll data [24:22] and other sources that are currently being deployed. [24:26] But I would say that it's this, right? [24:28] The farm all eventually transformed farming across the Corn Belt, [24:31] but it required the right conditions. [24:33] Congress can help create those conditions for AI. [24:35] The opportunity is very, very real, but so is the work that's required to do it. [24:40] So thank you, and I look forward to your questions. [24:41] You join? [24:47] And please, on behalf of the chairman, Dr. Sarah Myers-West, [24:50] you're welcome with your opening remarks. [24:53] Thank you. [24:54] Chairman Scott, ranking member Warren, and esteemed members of the committee, [24:58] thank you for inviting me to testify on this important set of issues. [25:03] My name is Dr. Sarah Myers-West, and I, along with my colleague Amba Koch, [25:06] lead the AI Now Institute. [25:08] We're an independent policy research center funded in 2017 [25:11] to study artificial intelligence technologies. [25:15] I also previously served as a senior advisor on AI at the FTC, [25:18] where my role focused on the effects of AI [25:21] as they were emanating across the broader economy. [25:24] There's a tremendous amount of hope articulated in the vision for AI, [25:29] that it will have transformative potential [25:31] that could lift our economy and improve the quality of life for all of us. [25:35] We're equally seeing a tremendous amount of pessimism from the public [25:40] about what this technology means for their lives and their livelihoods. [25:45] This divide is because there are skewed market incentives [25:47] that have shaped corporate AI development [25:51] and that are driving concentrated risk in the economy [25:54] that the public is being asked to absorb. [25:57] Importantly, many of these harms are already here, [26:00] and they're being felt by everyday people, [26:02] and they're going to continue out into the future, [26:04] especially if things go exactly as the industry intends. [26:07] My message to this committee is that we must put the needs of the public [26:11] back at the center of the AI economy. [26:14] Congress has an obligation to intervene before an economic crisis occurs, [26:19] starting by alleviating the burden that people are already feeling from AI, [26:24] exercising congressional oversight over AI firms [26:27] that have grown in size and in political influence [26:31] and stopping a bailout of the industry that is arguably already underway. [26:36] There are three core areas that I urge this committee to consider [26:40] as priorities for intervention. [26:43] First, addressing the risky and capital-intensive nature of AI investment. [26:49] AI companies are engaged in a race to the bottom, [26:52] one in which firms have exhibited an extraordinary willingness to court risk [26:56] that's not just borne by the firms themselves, [26:58] but by the wider economy and the public. [27:02] AI development is increasingly being financed [27:04] through complex and opaque instruments [27:07] that hinge on the private credit market [27:09] and are often kept off balance sheet [27:11] through the use of special-purpose vehicles. [27:13] We're also seeing circular spending deals [27:15] that hinge on the projected future revenues [27:18] of just a handful of AI companies, [27:20] and if those revenues fail to transpire [27:23] to the tune needed to justify this capital expenditure, [27:26] the house of cards could come falling down. [27:28] Second, the effects of an AI bubble [27:31] would be acutely felt by the American people. [27:34] It could wipe out a tremendous amount of American household wealth, [27:38] and this is because the economy at large [27:41] is already significantly exposed to AI risk [27:43] across several vectors, [27:45] ranging from increased market exposure for everyday families [27:48] to access to credit to job loss. [27:51] We don't have to wait for a financial collapse [27:53] to see these effects of AI play out across the economy. [27:56] Working-class Americans are particularly exposed [27:59] to an AI-driven market downturn. [28:02] For one, the private credit companies have made a casino [28:06] of American workers' financial security. [28:08] We're seeing a transfer of wealth from the public [28:11] to these AI firms, devaluation and dislocation already emerging [28:15] in the job market, and rising prices, [28:17] especially in the utility sector, [28:19] that could serve to depress consumer demand. [28:22] Importantly, many of these factors will remain [28:25] even in a scenario where there's no bubble burst. [28:27] The vision for success in AI hinges on the significant displacement [28:31] and devaluation of jobs across the economy. [28:35] I want to underscore if the vision for AI fails, [28:40] we could face a market correction. [28:42] If it succeeds, it looks like mass displacement, [28:44] and this is not a bargain any of us need to accept. [28:46] This is why it's important for Congress to stop the bailout [28:50] of the AI sector that's already in many ways underway. [28:53] AI firms have received extraordinary subsidies [28:56] from federal backing of loans to direct funding [29:02] to allocating federal lands for data center infrastructure. [29:06] And this is all before Congress has had the chance to meaningfully act [29:09] to protect the public and the economy from the risks posed by the industry. [29:13] The public's already aware of the risks that they're being asked to assume, [29:18] and this is fueling the generalized antipathy [29:21] that they're expressing towards AI. [29:23] If the end goal is to promote American innovation [29:26] and to build an economy that works for everyone, Congress must act. [29:31] It must start by focusing on measures that can alleviate the economic burdens [29:35] that are already being felt by the public, [29:37] by shedding light on AI industry financing structures [29:41] through the use of its oversight authority, [29:42] and most importantly, by committing to stop the bailout [29:45] before it deepens any further. [29:47] Thank you. [29:51] Thank you very much. [29:51] I appreciate your testimony today. [29:53] I'll start by asking the first questions. [29:56] Our economy will benefit immensely from American companies' leadership [30:00] in AI and advanced technology. [30:02] However, we received a letter just recently today from Anthropic [30:07] that Alibaba executed the largest known distillation attack [30:12] on Anthropic to date. [30:13] That means that our adversaries are willing to lie, cheat, and steal [30:17] to get access to the information necessary to make China the most dominant player [30:22] in the market, as opposed to the free market economy that we have [30:26] that produces the world's greatest thinkers and innovators. [30:29] I think that's a very important point because when China lies, cheats, and steals, [30:36] the American consumer pays the ultimate price for that. [30:39] When I'm in restaurants in Charleston and around the state, [30:42] one thing I hear a lot about is the cost of electricity, water usage. [30:47] There is this fear associated with higher prices in an economy [30:53] where it's very difficult for people to live paycheck to paycheck. [30:56] And so when you put together these amazing attacks from our adversaries, [31:02] the fear of our consumers, it puts us in a position where, [31:06] as a committee and as responsible leaders, [31:08] we need to have a better appreciation for how the costs can come down [31:13] and how to push back against the threats that could drive our cost up [31:18] and make us less relevant. [31:19] Mr. Flynn. [31:22] Senator, thank you for that question. [31:24] And I think you laid out the depth of the challenges really well. [31:27] When you think about the distillation attacks in particular, [31:31] that's a real challenge that they extend to which China will go. [31:37] To me, the answer is about where is the US AI stack? [31:41] Are we seeing it in Europe and the global south? [31:44] And I think it's worth saying that's not a guarantee. [31:47] The Spanish Ministry of the Interior just recently signed a significant deal [31:50] with Huawei, despite the fact that the US government over several years [31:53] has done a lot of education and advocacy going back to the 5G conversation [31:58] not that long ago about the dangers of having the cybersecurity risks [32:01] of having Huawei in their core infrastructure. [32:04] Still, it seems like we're having that conversation all over again. [32:11] Mr. Reinhart, chat GPT is socialized throughout all of our devices, it seems like. [32:17] And, of course, all the other ones are working really hard as well. [32:20] How do we help the average family in this country benefit from lower prices, more information? [32:30] I'll use this as an example. [32:32] Someone was looking for an air conditioning unit at home, [32:35] and they used these models to do research. [32:39] It would take several hours in 90 seconds. [32:43] I think we have a lot of opportunities to share with the American people the benefits, [32:49] the opportunities that come with this technology. [32:52] I'm not sure that the information is getting to the end user, [32:56] but certainly the benefits should. [32:58] Yeah, figuring out how to use technologies properly is something that is just a thankless task [33:04] and something that is very difficult to do. [33:06] I mean, I work with this every single day, and even I'm not completely sure of some of the capabilities. [33:11] But we do see a lot of people using, as you mentioned, [33:14] we see people using agents, as they're called, to figure out the best prices for airlines, [33:20] for a whole bunch of range of different products that they're buying. [33:23] And to me, that I think is a really prime benefit of these AI technologies, [33:29] is giving consumers choices and showing them what's actually happening in the marketplace [33:34] and giving them the ability to choose the better price and things that work for their families [33:38] and for themselves. [33:40] Excellent. [33:40] Mr. Feith, what is the single biggest bottleneck to maintaining U.S. leadership in AI right now? [33:53] Thanks, Senator. [33:54] I think it's compute broadly, which includes both the chips and the chip-making equipment, very crucially. [34:01] You said that in your testimony. [34:02] That's one of the reasons why I called on you to ask that question. [34:04] It helped me understand perhaps the larger perspective on export controls and the role it should play [34:14] in maintaining America's stronger position, especially against China. [34:19] But perhaps the part of the question I would like for you to answer in 25 seconds, by the way, [34:25] is how do we make sure that our allies don't have a back door, whether in the Middle East or Asia, [34:34] that goes to China and gives them access to the most powerful chips? [34:37] Now you have 17 seconds. [34:38] It's important. [34:40] There's a very big allied component, which partly reflects the strength that the U.S. [34:44] and our allies together have in the leadership in these key technology areas. [34:50] That's in semiconductor design and in the manufacturing equipment. [34:54] But getting allies, including the Dutch and the Japanese, especially on board, is really very important. [34:58] And we have only a mixed record of doing it over the last several years. [35:00] You did what I could not do in 23 seconds. [35:03] Good job. [35:04] Reagan member? [35:05] Thank you, Mr. Chairman. [35:07] So we need to address the serious national security risks of allowing advanced AI chips to end up in China. [35:15] Last week, the New York Times reported that NVIDIA's chips have been, quote, [35:21] sought by nearly every branch of the Chinese military. [35:26] And China's been getting its hands on this technology. [35:29] Federal prosecutors allege that billions of dollars of chips and servers have been unlawfully diverted to China during the Trump administration. [35:40] Now, I invited NVIDIA's CEO here to testify. [35:44] He has claimed there is no evidence of chip diversion and that the Chinese military doesn't want his chips. [35:54] But he will not show up here to repeat those claims under oath. [35:58] Mr. Fyfe, you served on President Trump's National Security Council until last year. [36:04] How worried should we be about these chips falling into the wrong hands? [36:08] Thanks, Senator. [36:10] I think very worried. [36:11] I mean, the implications are very significant in ways that you pointed out, in part military, domestic repression, also China's commercial strategic ambitions. [36:20] There's an important additional aspect, which is that if we sell them these chips, it actually helps China compete against us in third markets around the world, [36:28] helps them export their cloud, for example, and undermine the market for U.S. cloud exports all around the world when we're interested in otherwise maximizing the exports of the AI tech stack of the United States around the world. [36:40] So bad, bad, bad, bad is what I'm hearing you say. [36:43] Now, Congress has pending legislation to put a stop to that. [36:47] Senator Banks and I have introduced the AI Overwatch Act, which would ban the export of advanced AI chips to China. [36:55] Senator Kim and I have co-sponsored the Match Act, which would ensure China's not getting the equipment it needs to manufacture advanced chips on its own. [37:07] And Senator Cotton and I have co-sponsored the Chip Security Act, which would ensure that we can track when AI chips are smuggled. [37:17] Mr. Fyfe, do you think Congress should pass these three bills? [37:21] Yes, and kudos on the bipartisanship. [37:23] Good. [37:23] Mr. Chairman, there's broad support for these bipartisan bills. [37:28] I think we should mark these up, this legislation this month, and I hope you'll have a chance to agree on that. [37:35] Well, the gentleman woman from Massachusetts, you have the rest of your time and I'll talk about it. [37:39] No. [37:39] Okay. [37:41] So I want to turn, though, to how the AI boom is being funded. [37:48] AI companies are on track to spend a whopping $7 trillion on AI infrastructure by 2030. [37:54] And are turning to shadowy financing alternatives to meet their financing needs. [37:59] If the AI companies can't make payments on their huge debt loads, they could trigger massive losses across the financial system. [38:09] Dr. Meyers-West, if an AI bubble bursts, what would be the impact on families who may not have a single dollar of investment in an AI company? [38:21] Would they be hurt? [38:22] Dr. Meyers-West, the impact on them would be significant even if they haven't directly invested in AI because we could see compounding effects across people's retirement funds, their pension funds. [38:35] We'll see job layoffs and a potential credit freeze. [38:38] Dr. Meyers-West, the underlying conditions of our economy make us very ill prepared to deal with this kind of market exposure and our social safety net is not well prepared for these kinds of conditions. [38:49] So it could be a perfect storm that will be felt most acutely by people who are already struggling to make things work and would be irresponsible to look away. [38:56] Dr. Meyers-West, thank you for that answer. [38:59] You know, AI executives are already well aware of what's at stake and they are right now trying to lay the groundwork to leave American taxpayers holding the bag if their big bets go bust. [39:14] Open AI, for example, has pushed the Trump administration to, quote, lean in and assist the industry by, quote, de-risking AI expansion, including expanding federal tax incentives and loan guarantees for AI companies. [39:32] So I'd like to get our panel of experts on the record and just go down the line, one sentence or even one word. [39:40] Would you support the Trump administration bailing out big tech CEOs and AI companies? [39:48] Dr. Meyers-West? [39:51] I would not support a bailout of the AI industry. [39:53] Mr. Reinhardt? [39:55] Goodness, no. [39:56] I like that. [39:57] Mr. Fyfe? [39:58] Hard to generalize. [39:59] I would say there are some funds that come from places like the CHIPS Act that the Commerce Department still has and could use very productively. [40:06] I understand that. [40:07] But do you want to bail out AI if these bets go bad? [40:11] I think the CHIPS Act on on-shoring semiconductor manufacturing is a better bet. [40:14] Ah. [40:16] Mr. Flynn. [40:17] I would agree with David. [40:18] I would also say that I haven't seen any particular proposal. [40:22] I'm sorry? [40:22] I haven't seen any particular proposal but happen to take it back. [40:25] Happy to take it back. [40:26] I'm out of time, but I've got to say, trying to backpedal on the question of whether or not these guys should make trillion-dollar bets [40:33] and then be able to turn around and say to the American public, the taxpayers are going to bail us out. [40:39] We did that in 2008. [40:40] We shouldn't be doing it again. [40:42] Sorry, Mr. Chairman. [40:43] I apologize. [40:44] Senator Realms. [40:45] Thank you. [40:47] First of all, let me just thank you all for being here. [40:49] And second of all, I really appreciated the opening statements that all of you provided. [40:53] Mr. Flynn, I'm just going to begin with you. [40:56] Do you know of any bailouts requested by any of the current AI companies? [41:04] Not to my knowledge, Senator. [41:05] Does anybody know of any bailouts that are being requested by any of the AI companies? [41:12] See, here's the deal. [41:13] I think it's fair to say in this country, we recognize that artificial intelligence is here, and it's not going away. [41:21] The real question is, is China going to lead in AI development and take the opportunities, [41:28] or is the United States going to take the lead and be the place where the AI development actually occurs? [41:34] In the United States, we believe that it should be for the betterment of everybody [41:37] and that the developments are such that it will grow our economy. [41:41] Mr. Flynn, has the development of artificial intelligence in the United States added to our GDP? [41:46] Absolutely, Senator. [41:48] Mr. Fythe? [41:49] Yes, sir. [41:50] Mr. Reinhart? [41:51] Indeed, it has, yes. [41:53] Dr. Myers-West? [41:56] We've seen a number of places where the American public has already been feeling the pain. [42:01] You can't tell whether or not AI has actually developed or improved GDP growth in the United States? [42:07] The predominant factor has been the investment in AI infrastructures rather than AI itself. [42:12] So, you see, what we've got right now is in the United States, [42:15] we've done everything we can on a bipartisan basis to make sure that the United States does not fall behind. [42:21] If we don't do this correctly, AI will be used in the defense of our country, [42:26] but it will also be used by China in offensive operations. [42:30] Today, in the United States, we use it in both offensive and defensive operations. [42:35] We're better than our adversaries are right now. [42:37] That saves American lives long term. [42:40] But if we're going to do that correctly, we have to make sure that this is the place where those developments occur. [42:45] That means that the rest of the world, which is also a market for artificial intelligence [42:49] and the systems that literally people are going to use, [42:54] they're on the rails that are developed in the United States today. [42:58] We don't want those rails to be China's rails. [43:01] We don't want Huawei doing it. [43:02] We don't want the companies there doing it and taking advantage of it. [43:05] So, Mr. Flynn, what happens if an organization such as NVIDIA is not allowed or is restricted [43:13] from being able to market their products around the rest of the world? [43:16] Does that mean then that China then has an advantage in terms of selling to the rest of the world? [43:20] Senator, I think this is a core question of this hearing. [43:22] There's an inherent tension that needs to be addressed, [43:25] and this is why the role of BIS is so profoundly important, [43:28] not just on the export control side, on the enforcement side, [43:31] but their ability to issue licenses in a timely fashion [43:34] and to work with the regulated community so the information necessary [43:38] to ensure that the U.S. technology stack is being used around the world. [43:42] Mr. Feith, I really appreciated your background [43:44] and the fact that you've indicated that while we need to continue to lead the world, [43:50] there also needs to be the appropriate way [43:52] to make sure that we keep that away from other adversary countries such as China. [43:59] Today, would it be appropriate if we were to get the Commerce Department additional resources [44:03] to be able to monitor and to restrict that from getting into the wrong hands in China? [44:09] Yes, certainly. [44:10] Okay. [44:11] Let me run this by you again once on this thing here. [44:14] Is there any reason that we should be trying to chase any one of these organizations, [44:18] these hyperscalers, [44:19] any reason why we should chase them out of the United States [44:21] by trying to tax them at a higher rate than other businesses would be taxed? [44:25] No, I don't believe so. [44:27] Mr. Flynn? [44:28] No, Senator. [44:29] Is there any advantage at all to the American people [44:32] if another country, including China, [44:34] takes the lead on the development of artificial intelligence [44:37] in this once-in-a-generation opportunity to advance GDP in our country? [44:43] Certainly not. [44:45] Goodness, no. [44:46] I do not. [44:48] I fear a world where China is leading on AI. [44:50] Can you imagine a world in which if you use artificial intelligence correctly, [44:54] you can actually improve the quality of life for Americans? [44:58] We don't talk about it much. [45:00] But what happens as we integrate biomedical research [45:03] and the development of artificial intelligence, [45:05] being able to make decisions at a very, very rapid pace, [45:08] Mr. Flynn, does that make a difference for the quality of life [45:11] for the American families? [45:12] Absolutely, Senator. [45:14] Mr. Feit, do you agree with that? [45:16] Yes, sir. [45:17] Mr. Reinhardt, you've been kind of quiet today, [45:18] but would it be fair to say that there are real opportunities [45:21] for the American public to benefit from the development [45:24] of artificial intelligence in the United States rather than in China? [45:28] Yes. [45:29] Yes, very clearly. [45:30] And I would say one thing I saw recently, [45:32] an individual had a very rare complication to a heart medication. [45:37] And by putting it into CHAP-GBT, [45:39] he realized that he needed to switch his heart medication. [45:42] So I think that these small stories are actually incredibly important [45:44] for individual lives. [45:45] Thank you. [45:46] Look, I'm running out of time, but here's the deal. [45:49] Artificial intelligence is here. [45:50] It's not going away. [45:52] We've got to be able to do the right types of guidelines [45:56] for the development of it, [45:58] but we have to do everything we can to keep that development here [46:01] in the United States, both for economic development, [46:03] but for the quality of life that it proposes for the people [46:07] that live in this country. [46:08] Thank you, Mr. Chairman. [46:09] Thank you, sir. [46:10] Senator Reid. [46:10] Thank you very much, Mr. Chairman. [46:13] Thank you to all the witnesses. [46:16] Mr. Feith, there is concern. [46:19] I heard a bit of Senator Roundt's testimony [46:23] about providing China with H-20 and H-200 chips from NVIDIA. [46:31] And one reason, and this is what I've heard from national security experts, [46:37] is China is very rapidly and very adeptly using those chips [46:42] to expand its very sophisticated and indeed weaponized chips. [46:48] Is that accurate? [46:51] Yes, sir. [46:52] I think we've seen valuable testimony from the U.S. military and elsewhere on that. [46:56] Now, they are, China is promoting, and promoting is an understatement, [47:04] the Huawei chips and other chips. [47:08] They are not, I don't think, bringing the H-20, H-200 chips in [47:13] to displace their own chips. [47:15] On the contrary, I think they're doing all they can [47:17] to promote their own Huawei chips and others. [47:20] Is that right? [47:21] That's right. [47:22] They want very much to indigenize, and they want to gain [47:26] so that they can not only satisfy their domestic market with Huawei chips, [47:30] which right now they're way behind in their ability to do, [47:32] but their ambition is to do so. [47:34] And then beyond that, they want to then compete to export all around the world [47:37] against U.S. chip makers and cloud companies. [47:40] So is it fair to say that what they really want from H-20 and H-200 chips [47:45] is to get the most advanced American technology they can [47:50] to incorporate in their chips and also to incorporate in their sophisticated weapon systems? [47:58] Yes, and I would say that there are other chips, including the Blackwells, [48:02] that they've demonstrated a major interest in getting as well, [48:04] even though they're not licensed for export, like through smuggling. [48:07] So the president's decision to allow them to get H-20 and H-200 chips, [48:16] what would you consider the rationale to be if, in fact, [48:21] they're using it to exploit our technology, not for any other reasons, perhaps? [48:30] Well, I think the administration stated that their calculation was that [48:35] these are not the leading-edge chips, they are second- or third-level chips, [48:41] and that sending them to China would therefore not give China the best, [48:46] but could complicate China's ambitions to indigenize. [48:50] My sense was that the calculations probably cut the other way, [48:54] but I do think it's important that the administration has been very clear. [48:57] The president has said repeatedly that the principle is to deny China the leading-edge chips, [49:01] deny them Blackwells. [49:02] That's been maintained. [49:05] I think it's important. [49:06] It's important to follow it up with robust enforcement, [49:09] including around smuggling and on possible loopholes. [49:11] No, I think that makes sense. [49:13] But I think, as you point out, the H-20 chips, et cetera, [49:21] are not going to be indigenous to China. [49:23] They don't want American chips to be dominant in their market, [49:27] and they are exploiting these chips, [49:30] and they may not be the most exquisite chips we have, [49:33] but frankly, they appear to be better than what the Chinese have, [49:37] and they're weaponizing them. [49:39] Is that fair? [49:40] That's right. [49:40] And I think part of your question also points to this question of, [49:45] you know, what is China really about? [49:46] Is it about simply maximizing their compute access [49:50] and potentially maximizing business for U.S. exporters, [49:54] or is it about indigenizing and replacing? [49:56] And I think China's ambitions in semiconductors are, [49:58] as they have been in other industries, [50:01] they will only take U.S. technology [50:02] as long as it takes them to indigenize and then overtake us. [50:07] We shouldn't think that we can addict them to these U.S. exports. [50:10] Thank you very much, Dr. West. [50:13] Dr. West, we have to think seriously about the impacts, [50:19] particularly in employment impacts, [50:21] of the wide-scale introduction of AI. [50:24] I know the former Secretary of Commerce, [50:27] Senator, Congress Secretary, excuse me, Raimondo, [50:31] wrote a very good editorial in the New York Times [50:35] indicating we're not ready yet for the impact. [50:39] Could you comment on that? [50:40] Yes, I mean, I think we're already seeing, you know, [50:44] firms narrating layoffs as due to AI, [50:48] whether or not the technology is, in fact, the causal factor. [50:51] And here, I think it's critical that workers [50:54] who are sort of on the front lines of experiencing [50:57] the impact of AI, whether that's being laid off [51:00] or their jobs being devalued and their wages depressed, [51:04] be positioned to have a crucial say [51:06] in the terms on which AI is deployed in their workplaces. [51:10] Thank you very much. [51:11] Janelle, you've got my second. [51:15] Thank you. [51:16] I'll give it to the ranking mover for five seconds. [51:18] I just want to say that when I was talking through the bills, [51:21] I failed to mention that the lead sponsor on Match [51:24] is Senator Ricketts, and it was just an oversight. [51:27] I apologize. [51:28] It's a terrific bill. [51:29] Senator Kennedy, the floor is yours, sir. [51:38] Mr. Feith, is that how you say your name? [51:41] Yes, sir. [51:45] I continue to read and hear that we have to win the AI race. [51:56] What does that even mean? [52:00] Generally, we heard that with 5G. [52:04] Generally, we hear that when somebody in the private sector says, [52:09] we're behind China, give us some money. [52:14] If we are ahead, eventually China's going to catch up, [52:21] and then China gets ahead, and we catch up with them. [52:29] Ultimately, and there's no finish line here, [52:33] where the first one who crosses gets a gold medal. [52:38] This stuff keeps evolving. [52:41] Aren't we ultimately going to have to sit down with China [52:46] and reach some sort of compromise [52:50] in terms of a modicum of cooperation? [52:56] Well, I think, Senator, I think the questions [52:59] about whether this race vocabulary and metaphor [53:04] are exactly right are very well taken. [53:07] I think it's inexact at best, [53:12] but there are comparisons possibly [53:16] with previous generations of strategic technology [53:20] like nuclear weapons, [53:21] where I think maintaining that the U.S. and our allies [53:25] have a stronger foundation of strategic technology, [53:28] including as it needs to be innovated across generations [53:31] and including as, you know, [53:33] after the Soviets got a nuclear weapon in their time [53:36] several years after us, is extremely important. [53:38] And the fact that this is commercially innovated... [53:40] I got it. I'm sorry. [53:42] I don't mean to be rude, [53:43] but I don't have time for history. [53:47] Dr. Winston, I want to make sure you're heard. [53:56] Well, before I come to you, [53:58] let me ask you, Mr. Reinhardt, [54:01] do you think we ought to have state-owned AI companies? [54:06] Are you a fan of the state-owned companies in China? [54:12] No, not at all. [54:14] Whether the federal government takes a big stake [54:16] and runs the show? [54:17] No, I do not. [54:18] We can't even deliver the mail, can we? [54:22] Especially in my neighborhood, yeah. [54:23] And the stuff has an address right on the front, doesn't it? [54:26] Yeah, I think that we can be smart about how we do grant programs. [54:32] I mean, there's ways that we can be smart about the... [54:34] Grant programs? [54:35] Well, just the various, you know, we've had, [54:38] as we've talked about, some of the CHIPS Act, [54:40] parts of the CHIPS Act program are still out. [54:43] I understand there are alternatives to grants. [54:45] They're called loans. [54:47] Of course, yes. [54:48] Okay. [54:49] Yeah. [54:49] I'm just saying that I think there's a lot of, [54:51] there's lots of parts of the federal government [54:53] that we can think about being smart about our investments in AI, [54:56] and that doesn't necessarily require any sort of ownership. [55:00] Well, look, I think the people developing AI, [55:06] this is true worldwide, [55:07] but it's certainly true in the United States. [55:09] These are brilliant people. [55:10] Right now, they're spending money to try to be the biggest [55:21] and the best and the wealthiest. [55:25] And I don't mean disrespect when I say this. [55:27] I'm proud of them as American entrepreneurs, [55:30] but I trust them like I trust the rest stop bathroom. [55:36] I remember when one of the leaders in another hearing about it, [55:42] he asked him, I said, [55:43] man, you must make a lot of money. [55:46] And he said, oh, no, no, no, I do this for the goodness of the people. [55:51] I only get my health insurance paid. [55:54] Wrong. [55:57] So, Dr. West, you make a good point, [56:00] but I guess my question is, what do you expect us to do? [56:05] It's here. [56:07] And what do you, people are going to lose their jobs. [56:11] I hate to see that. [56:12] I might lose mine. [56:15] But what are we supposed to do about that? [56:21] I think it's a great question [56:22] because there's this sort of generalized environment [56:24] that this is inevitable [56:25] and we all have to just kind of deal with the consequences. [56:28] Yes, ma'am. [56:29] I understand the problem. [56:30] I'm looking for, I admire the problem. [56:33] I'm looking for a solution. [56:34] Tell me what we think we ought to do [56:35] when somebody loses their job as a result of AI. [56:39] In the next 10 seconds or so. [56:41] Oh, wow. [56:42] I think we need strong accountability measures [56:45] that restrain the actions of this industry [56:49] that are sort of experimenting in the wild [56:51] with the release of these tools. [56:52] There's a request that tells me nothing. [56:55] I'm sorry. [56:56] I'm way over. [56:58] But you're a good man for letting me go over it. [57:00] Yes, sir. [57:01] Well, I just wanted to associate myself [57:03] with your comments about the rest stop bathrooms. [57:08] There's not a lot of confidence in those. [57:10] You got to be careful. [57:12] The coffee's lousy, too. [57:13] Yes, sir. [57:14] Well, I hadn't had that experience. [57:16] Especially in Maine, evidently. [57:18] Senator Warner. [57:19] Well, I'd like to invite my colleague [57:22] from South Carolina and Louisiana. [57:25] A certain former governor of Virginia [57:26] made it a top priority to improve the quality [57:29] of all the rest stop bathrooms in the Commonwealth. [57:32] And we welcome you to use our facilities [57:35] at your discretion. [57:39] Let me answer, I think, you know, [57:41] I think what Senator Kennedy asked [57:42] is a really good question. [57:44] And I've talked to every one of the big heads. [57:47] I spend more time on AI, probably next to Mike Rounds [57:49] of anybody here. [57:50] And there is no macro answer. [57:52] Four years ago, we thought, [57:53] well, let's make everybody learn to code. [57:55] That was well-intentioned, but not the right answer. [57:58] Frankly, one of the things we ought to do [58:00] at a macro level is in high school, [58:03] junior and senior year, start convincing people [58:05] that if you're going to major [58:07] in business, business administration, economics, [58:10] which is the biggest majors in public universities, [58:12] and those are the jobs that are going to be [58:13] least most disappearing, [58:15] we ought to convince a heck of a lot more people [58:16] to go into health care, [58:18] where there will be jobs [58:21] for some time to come. [58:23] I'd also say I think the analogy on the race [58:25] is a little wrong, too, [58:26] and I absolutely agree with Mike. [58:29] We've got to stay ahead [58:31] on the leading frontier models. [58:34] But the fact is, even right now with us ahead, [58:37] the Chinese open source models are 10x cheaper, [58:40] use a lot less tokens, [58:41] a lot less electricity, [58:43] and while we're leading, [58:45] most of our smaller entities [58:47] are already using Chinese models. [58:50] And I hope the panel has seen, [58:53] there was a Booz Allen study out recently, [58:55] I think it's quite good, [58:56] and I recommend it to all my colleagues, [58:58] that if you ask the Chinese models [59:01] a question in English or French, [59:04] anything other than Mandarin, [59:05] you get a different answer. [59:08] And if you happen to note [59:10] that I'm working on a government project, [59:12] you get an answer [59:13] that comes with flaws built in. [59:16] And so the idea that China, [59:18] having been in the telecom business [59:20] and seen this movie with Huawei, [59:23] this is nothing compared [59:25] to what's happening in Chinese models. [59:26] Where I disagree with some of my colleagues, [59:29] the idea that we're going to just do [59:30] pedal to the metal regardless of outcomes, [59:33] I think most of us agree we screwed up [59:34] by not putting some guardrails on social media. [59:37] This is peanuts compared to where. [59:41] So the way I'm looking at this, [59:42] and I apologize, [59:42] I will try to get to a question, [59:43] is I think we need to realize, [59:45] I don't think moratoriums on data centers make sense, [59:47] but I sure as heck think [59:48] making sure there's no misuse of power, [59:51] water, backup generators, setbacks, [59:54] non-disclosure agreements. [59:56] So communities, [59:57] what they're getting into [59:57] is an enormous amount of sense. [59:59] And they have to pay part of the share. [1:00:01] And there's a bunch of ideas [1:00:03] around how we address that. [1:00:05] I'm looking at an idea [1:00:06] that doesn't raise their taxes necessarily, [1:00:09] but since the benefit will be deferred, [1:00:11] maybe we ought to defer [1:00:12] some of the accelerated depreciation [1:00:14] for a couple of years [1:00:15] so we can gain some revenue [1:00:17] to help on the retraining [1:00:18] as we decide what to retrain. [1:00:21] We think about social media, [1:00:22] social media to your kids [1:00:23] is nothing compared [1:00:24] to what a chatbot can do your teenage son. [1:00:27] Where I think back four years ago, [1:00:28] we had every one of the AI companies [1:00:30] promise in elections, [1:00:32] we will never do deep fakes. [1:00:34] Bull, they're out there. [1:00:36] They could correct that. [1:00:37] We do need to have them put pedal to the metal. [1:00:41] And mythos, thank God, [1:00:43] it was anthropic. [1:00:46] When the head of the NSA [1:00:47] and cyber commands came and said, [1:00:49] this tool broke into [1:00:50] almost all of our classified systems, [1:00:53] not in weeks, but in hours. [1:00:56] And one of the things [1:00:57] I don't know how we do, [1:00:58] and I know there's a big debate [1:01:00] in the administration, [1:01:01] but we're not going to get this fixed [1:01:03] if we have a less ethical CEO [1:01:05] in terms of simply voluntary pre-testing. [1:01:10] Now, how we get to the right incentives, [1:01:12] it doesn't have to be 90 days, [1:01:14] it could be 30 days. [1:01:15] And I do think the dislocation, [1:01:18] I'll make a bet with any of you here. [1:01:21] John, we're at 9% [1:01:22] recent college graduate unemployment. [1:01:24] I will bet you dollars of donuts right now, [1:01:27] it'll go to 30% by 2028. [1:01:32] And that will have a double whammy, [1:01:34] not just the age cohort, [1:01:35] but you're going to have [1:01:36] a lot of pissed off parents and grandparents [1:01:37] who just spent $250,000 on an education. [1:01:40] And we hear a lot of talk about, [1:01:43] you know, AI adjacent jobs. [1:01:45] I do believe they will come, [1:01:47] but we need a whatever, [1:01:49] the Democrats will call it a New Deal program. [1:01:52] Maybe we can call it a Reagan era program. [1:01:55] Something that says for five years [1:01:56] how we get through this gap. [1:01:57] And we've got to have the industry help us. [1:01:59] If it's simply a government training program, [1:02:01] we'll screw it up. [1:02:03] So Mike, I was going to ask you on CISA [1:02:06] and the ISACs, [1:02:08] which I think need to be refunded. [1:02:10] Mr. Flynn, I think you've even mentioned that. [1:02:12] And by the way, [1:02:12] you've also supported the bill [1:02:13] that Senator Rounds and I have. [1:02:14] Thank you for that. [1:02:15] Quick comment on the ISACs [1:02:16] and then I'll get out of your hair. [1:02:17] Profoundly important, Senator. [1:02:18] And thank you for so many years [1:02:20] of excellent work on cybersecurity, [1:02:22] elevating that issue set, [1:02:24] which is wonky [1:02:25] and not everybody wants to talk about it. [1:02:26] The ISACs are a huge piece of this equation, [1:02:29] making sure that the highest leveraged [1:02:30] cybersecurity players, [1:02:31] their information on vulnerabilities, [1:02:33] other threat information, [1:02:34] gets down to players with fewer resources [1:02:36] is critical to making sure that we're prepared. [1:02:38] When I get that response, [1:02:39] I'm going to leave. [1:02:40] Thank you all. [1:02:40] Apologize not getting any of the others. [1:02:42] Thank you. [1:02:42] Senator Ricketts. [1:02:43] Just want to remind my colleague from Virginia [1:02:45] that betting is no longer allowed. [1:02:50] Senator Ricketts. [1:02:52] Thank you. [1:02:53] Thank you to all of our panelists [1:02:54] for being here today. [1:02:55] Communist China is trying to dominate AI. [1:02:59] They're trying to do it [1:03:00] for a variety of reasons, [1:03:02] but one of the biggest reasons [1:03:03] is they think this is how [1:03:05] they can dominate us militarily, [1:03:07] whether it's through dorm swarms, [1:03:11] hypersonic missiles, [1:03:13] attacking us through cybersecurity, [1:03:15] that sort of thing. [1:03:16] And semiconductor manufacturing equipment, [1:03:19] or SMEs, [1:03:20] is the ultimate choke point [1:03:22] in this competition. [1:03:25] Many people are familiar [1:03:26] with the amazing lithography machines, [1:03:28] EUV and DUV, [1:03:30] for printing features on chips. [1:03:33] Yet SMEs is much more than that. [1:03:36] It also includes hundreds [1:03:38] of other advanced technologies [1:03:40] like testing machines, [1:03:42] etch tools, chemicals, [1:03:44] and specialized design software. [1:03:47] Chips, models, and applications [1:03:48] all flow downstream [1:03:49] from this hardware. [1:03:52] Despite pouring billions of dollars [1:03:53] into domestic SME initiatives, [1:03:55] Beijing still hasn't indigenized [1:03:57] these highly complex technologies. [1:04:00] That's why if you are going to control [1:04:02] one part of the semiconductor supply chain, [1:04:05] it should be the SMEs. [1:04:06] Mr. Feith, [1:04:07] do you agree that export controls [1:04:08] on SMEs are the most effective way [1:04:10] to leverage the limit [1:04:11] of the PLA on its AI capabilities? [1:04:15] Yes, Senator. [1:04:15] Thanks for your leadership on that. [1:04:17] The Trump administration [1:04:18] pioneered using export controls [1:04:20] on SMEs [1:04:21] to blunt China's chip ambitions. [1:04:23] Key allies eventually followed, [1:04:25] but our lead has dangerous gaps, [1:04:27] and those still remain. [1:04:29] The controls are narrow [1:04:30] and less restrictive, [1:04:31] and they have loopholes. [1:04:33] Foreign companies [1:04:33] are backfilling the Chinese market [1:04:35] with critical dual-use tools. [1:04:37] This harms American workers, [1:04:38] risks our compute lead, [1:04:40] and fundamentally undermines [1:04:41] our collective national security. [1:04:43] As has been mentioned before, [1:04:45] and thank the ranking member [1:04:46] for mentioning it as well, [1:04:48] I introduced the Match Act [1:04:49] with Senator Kim. [1:04:50] Not only in addition, [1:04:52] I'll throw out Senator Risch [1:04:53] and Senator Schumer [1:04:53] also co-sponsors on that bill as well. [1:04:56] And I would encourage this committee, [1:04:58] as the ranking member did, [1:04:59] to do a markup on this bill [1:05:01] so that we can move on this [1:05:02] as quickly as possible. [1:05:04] Our bill prioritizes diplomacy first, [1:05:06] giving the state and commerce departments [1:05:08] a clear window to negotiate [1:05:10] with allied governments [1:05:11] and get the alignment, [1:05:12] but it also establishes [1:05:13] a firm deadline. [1:05:14] If our partners refuse to align, [1:05:16] the U.S. must act [1:05:20] to stop allied firms [1:05:21] from weaponizing this technology [1:05:24] and giving it to our adversaries [1:05:26] like Beijing. [1:05:27] Mr. Feith, [1:05:28] do you agree that weaker allied export [1:05:30] and servicing standards [1:05:31] on choke points for SMEs [1:05:33] create an unacceptable vulnerability [1:05:35] in our strategy [1:05:35] to contain China's AI ambitions? [1:05:38] Absolutely. [1:05:38] As you said, [1:05:39] the Match Act looks to build [1:05:40] on the first Trump administration's [1:05:44] export controls on EUV lithography, [1:05:46] which have been by far [1:05:47] the most effective [1:05:48] and crucial across years [1:05:50] and demonstrate the enormous benefit [1:05:52] that we can get [1:05:52] from these export controls. [1:05:54] And I think expanding that [1:05:54] in the way that the legislation points to [1:05:56] to make sure that we are aligned [1:05:57] with allies as much as possible [1:05:59] and that we have additional [1:06:01] lithography technologies [1:06:02] like DUV immersion covered [1:06:04] is really very important. [1:06:06] The Match Act provides [1:06:07] a clear timeline for diplomacy, [1:06:09] but we cannot negotiate indefinitely [1:06:10] while Communist China [1:06:12] continues to advance. [1:06:13] If our allies failed [1:06:14] to match our prohibitions [1:06:15] within the bill's time frame, [1:06:17] do you support applying [1:06:17] foreign direct product rule [1:06:19] to close foreign compliance gaps? [1:06:22] Mr. Feith? [1:06:22] Yes, and I think [1:06:23] that reflects the record. [1:06:24] I mean, it was the ability [1:06:25] to do that and to negotiate [1:06:27] with that possibility [1:06:28] that has gotten us [1:06:29] the best cooperation we've had [1:06:31] over the last eight or so years [1:06:32] with the Dutch. [1:06:34] We do want to negotiate, [1:06:35] but we also have to have a hammer, right? [1:06:37] Certainly. [1:06:38] If we don't act now [1:06:39] to close these loopholes, [1:06:40] we run the risk [1:06:41] that Beijing indigenizes [1:06:42] and dominates SME [1:06:43] just like they did for EVs [1:06:45] and critical minerals. [1:06:47] Do you agree that it is critical [1:06:48] that export control loopholes [1:06:50] are closed as soon as possible? [1:06:52] Yes, sir. [1:06:52] Do you know the reason [1:06:53] why we should get the markup on this? [1:06:56] China's civil military fusion strategy [1:06:59] ensures commercial breakthroughs [1:07:00] immediately advance the PLA [1:07:03] and it eliminates the boundary [1:07:05] between civilian technology [1:07:07] and military end uses. [1:07:08] By law, [1:07:09] Beijing asks companies [1:07:11] must hand over any technology [1:07:12] to the military. [1:07:14] The Pentagon understands [1:07:15] this is a massive challenge. [1:07:16] It just expanded a list [1:07:17] of Chinese military companies, [1:07:19] the 1260H list, [1:07:23] and added several prominent [1:07:25] consumer technology titans. [1:07:27] The list restored [1:07:27] restored memory chip makers, [1:07:30] CXMT and YMTC, [1:07:33] but it also added three [1:07:34] of China's top AI builders, [1:07:36] Alibaba, Baidu, and Tencent. [1:07:39] Because of this civil military confusion, [1:07:42] civil military fusion, [1:07:44] it is impossible to track [1:07:45] or control any technology [1:07:46] once it enters technology. [1:07:48] That's why we have [1:07:49] some of these other things [1:07:50] like the Chip Security Act [1:07:51] to mandate location [1:07:53] verification technology [1:07:54] as a key part [1:07:55] in managing this challenge. [1:07:57] We need to make sure [1:07:58] that we have this. [1:07:59] Mr. Feith, [1:08:00] if China's military [1:08:01] civil fusion economy, [1:08:04] do you support [1:08:06] the Match Act's [1:08:06] appropriate countrywide controls [1:08:08] as a choke point on SMEs [1:08:10] and its efforts [1:08:11] to apply these controls [1:08:12] to other affiliates [1:08:13] and subsidiaries? [1:08:14] Yes, certainly, Senator. [1:08:15] And I'll just say quickly, [1:08:16] on the 1260H list [1:08:17] and Alibaba, [1:08:18] it speaks to the question [1:08:19] that Chairman Scott [1:08:20] also led with, [1:08:21] with respect to Alibaba [1:08:23] and the anthropic distillation. [1:08:25] And having them identified [1:08:26] as military-affiliated companies [1:08:28] given military-civil fusion [1:08:30] is very important. [1:08:31] And we've got export control rules [1:08:33] that come into play [1:08:34] with that designation [1:08:35] that are also valuable. [1:08:37] Great. Thank you. [1:08:40] Senator Van Hollen. [1:08:42] Thank you, Mr. Chairman. [1:08:43] Thanks for holding this hearing. [1:08:44] Thank all of you [1:08:45] for your testimony today. [1:08:48] We're all trying to get [1:08:49] our arms around [1:08:50] a very rapidly moving [1:08:52] and developing AI universe [1:08:54] with increasing capabilities. [1:08:56] We recognize that [1:08:58] these capabilities [1:09:00] can bring benefits [1:09:01] to Americans [1:09:01] in areas like medical research. [1:09:04] And we have embraced innovation, [1:09:07] and we should continue to do so, [1:09:09] but we should make sure [1:09:10] we have the appropriate guardrails [1:09:12] in place, in my view. [1:09:13] As we do that, [1:09:15] millions of Americans [1:09:17] are asking themselves [1:09:18] what this means for their jobs [1:09:19] and their ability [1:09:20] to earn a living [1:09:21] through hard work. [1:09:24] There are questions [1:09:25] about whether or not AI [1:09:27] will replace humans [1:09:28] in making critical [1:09:29] military decisions. [1:09:31] I think that would be a mistake. [1:09:35] We also have AI agents, [1:09:37] and are they going to replace [1:09:39] individual decision-making [1:09:41] in many areas? [1:09:42] So lots of questions. [1:09:43] I'm glad we're having this hearing. [1:09:46] One area, of course, [1:09:47] is the question [1:09:48] of worker displacement. [1:09:49] That's an area of concern. [1:09:50] I've been looking at legislation [1:09:53] and proposed it in the past [1:09:54] before the questions of AI arose [1:09:57] with respect to how we can deal [1:09:58] with chronic long-term unemployment. [1:10:01] And I worry that that could become [1:10:04] an even more acute problem [1:10:06] going forward. [1:10:08] So, Dr. West, [1:10:10] if you could just speak briefly [1:10:11] to what you see [1:10:12] as the risks of workers [1:10:14] in terms of job displacement, [1:10:18] a lot of different theories [1:10:20] and warnings, [1:10:21] but where do you think [1:10:22] we are right now? [1:10:24] And would it be useful [1:10:25] to begin now [1:10:26] to look at how we can prevent [1:10:28] chronic long-term unemployment [1:10:30] that is at least part [1:10:31] and part caused by AI? [1:10:34] I think that's exactly right. [1:10:36] We are already seeing indicators [1:10:39] that workers are feeling [1:10:40] the experience of job displacement, [1:10:44] whether or not AI is at the root cause. [1:10:47] AI is sort of being attributed [1:10:48] as the underlying factor. [1:10:51] And our social safety net [1:10:53] is just not well set up right now [1:10:54] to be able to kind of catch people [1:10:56] when they experience that displacement. [1:11:00] And so I think focusing there first [1:11:02] and making sure that we design [1:11:05] our unemployment, [1:11:08] our other healthcare, [1:11:09] like the other elements [1:11:10] of our social safety net [1:11:12] that will help us contend [1:11:14] with the potential future impacts [1:11:17] on the workforce [1:11:17] is an important place to start. [1:11:19] No, I appreciate that. [1:11:20] And I think it's, as you say, [1:11:22] it's not too early [1:11:23] to really give that [1:11:24] some serious thought. [1:11:25] I want to turn briefly [1:11:27] to a question [1:11:27] that I heard Senator Warren [1:11:28] raise in her opening statements. [1:11:30] And Senator Ricketts [1:11:31] just alluded to it, [1:11:32] as others have, [1:11:33] which is the issue [1:11:34] of our export controls [1:11:36] on very high-end AI technologies [1:11:39] and chips. [1:11:41] I think we would all agree [1:11:43] that AI can be significant [1:11:48] in adding to military capabilities. [1:11:50] Is there anybody who disagrees [1:11:51] with that statement? [1:11:53] And so I have been concerned [1:11:55] about the loosening [1:11:57] of export restrictions [1:11:58] to some of our chief military adversaries, [1:12:02] like China. [1:12:04] There was just a New York Times story [1:12:07] the other day [1:12:07] that reported [1:12:08] that the Chinese military [1:12:09] has sought out high-end chips [1:12:10] from NVIDIA. [1:12:12] And President Trump [1:12:13] has seemed eager [1:12:14] to compromise national security [1:12:16] to grant favors [1:12:18] to friends and cronies. [1:12:21] NVIDIA CEO, Jensen Huang, [1:12:23] wants to sell more chips to China. [1:12:26] Like any company, [1:12:27] he wants to sell more [1:12:27] for his shareholders. [1:12:29] And he has been cozying up [1:12:30] to President Trump. [1:12:31] He gave money [1:12:32] for Trump's ballroom project. [1:12:35] He went to a million-dollar-a-plate dinner [1:12:36] at Mar-a-Lago. [1:12:38] And on and on. [1:12:40] And sure enough, [1:12:41] President Trump announced licenses [1:12:43] for NVIDIA to sell chips to China, [1:12:45] the H-20 and the advanced H-200 series. [1:12:48] And again, as we've all acknowledged, [1:12:51] these have military applications. [1:12:53] So we need to be careful. [1:12:56] Dr. West, what can we do? [1:12:59] And others may want to answer as well. [1:13:00] What can we do [1:13:01] to prevent this continued diffusion [1:13:04] of high-end chips [1:13:07] and AI technologies [1:13:08] that could come back to bite us [1:13:12] when they're deployed [1:13:13] by other militaries? [1:13:16] I think that you're right [1:13:17] to point out the chip problem. [1:13:18] I think what I would add to it [1:13:20] is, in fact, [1:13:21] that there is a broader challenge [1:13:23] faced by this sort of focus [1:13:26] on redeploying commercial [1:13:27] off-the-shelf AI models [1:13:29] in national security contexts. [1:13:31] We've seen the ways [1:13:32] that they have tremendous amounts [1:13:34] of security flaws [1:13:35] and vulnerabilities. [1:13:37] And I think it's important [1:13:38] that we also be addressing [1:13:39] the potential integration [1:13:41] of those vulnerabilities [1:13:41] into the critical infrastructures [1:13:43] in which these systems [1:13:45] are being deployed. [1:13:46] I really worry about the approach [1:13:48] of, you know, [1:13:49] using only voluntary measures [1:13:51] to test and validate [1:13:53] and look for these vulnerabilities [1:13:54] in these systems. [1:13:56] Appreciate it. [1:13:56] Anybody else want to weigh in [1:13:57] on this? [1:13:57] Thanks. [1:13:59] Senator, I think it's [1:13:59] a fantastic question. [1:14:01] BIS, strengthening BIS, [1:14:03] in short, export controls [1:14:05] are only going to be effective [1:14:07] if the Bureau, [1:14:08] if the agency charged [1:14:09] with doing that work [1:14:10] can effectively administer [1:14:12] those programs. [1:14:13] And right now, [1:14:14] I think we've seen, [1:14:15] at least what I've heard [1:14:16] from ITI companies, [1:14:17] is a real challenge [1:14:18] in BIS's ability [1:14:20] to administer the export controls [1:14:21] that are on the books. [1:14:22] Thank you. [1:14:23] I'm glad you raised that. [1:14:24] Thank you. [1:14:25] Senator Moreno. [1:14:26] Thank you, Mr. Chairman. [1:14:27] Thank you for having this hearing. [1:14:29] Just a quick lightning round [1:14:30] of questions. [1:14:31] Do you think the time [1:14:33] is come and gone [1:14:34] for data centers [1:14:35] to receive any type [1:14:37] of government incentive? [1:14:38] I'll lay that up for you, [1:14:39] Dr. West. [1:14:40] I'm going to assume no. [1:14:41] I think that they've received [1:14:42] enough incentive. [1:14:43] Right. [1:14:43] So no more. [1:14:44] So we could just say [1:14:45] no more, right? [1:14:46] We could just say [1:14:46] any city, county, state [1:14:49] does not need to give [1:14:50] any form of tax incentive [1:14:52] to any company [1:14:53] building a data center. [1:14:54] Yes or no? [1:14:56] I think these need [1:14:56] to be responsive [1:14:57] to the local communities [1:14:58] in which they're being deployed [1:14:59] and what we're hearing [1:15:00] from those communities [1:15:01] is no. [1:15:02] So no more money. [1:15:03] Mr. Reinhardt? [1:15:04] Generally speaking, [1:15:05] I'm not a big fan [1:15:07] of tax carve-outs, [1:15:08] you know, [1:15:08] broad base and low rates. [1:15:10] But I do think... [1:15:10] Just pay like anybody else. [1:15:11] Yeah. [1:15:11] Yeah. [1:15:13] That's right. [1:15:14] Sounds right. [1:15:16] The average American [1:15:19] interacts with a data center [1:15:21] more than 100 times a day. [1:15:23] Totally understand [1:15:23] the conversation [1:15:25] and the challenges [1:15:28] that constituents [1:15:28] may face in this. [1:15:30] And I think that [1:15:30] is incumbent [1:15:31] on ITI member companies [1:15:33] and our industry [1:15:34] to do a better job [1:15:35] in educating [1:15:36] local populations. [1:15:37] Data centers support [1:15:38] 5.5 million jobs [1:15:39] in America in 2024. [1:15:40] So there are benefits [1:15:41] to local communities [1:15:42] and it's on us [1:15:43] to make sure [1:15:44] that those communities [1:15:44] know that. [1:15:45] All that's great. [1:15:46] So do they get tax incentives, [1:15:47] yes or no? [1:15:50] I'm sure the answer is yes. [1:15:51] But they shouldn't, correct? [1:15:53] You agree that [1:15:53] that time is gone. [1:15:54] We should not be giving [1:15:55] handouts to companies [1:15:57] that are building data centers. [1:15:58] Would you agree? [1:15:59] Not necessarily. [1:16:00] Data centers are [1:16:00] an incredible part of... [1:16:02] How about car dealers? [1:16:03] Should they get tax incentives? [1:16:06] They employ thousands [1:16:07] of people too. [1:16:08] Absolutely. [1:16:09] But my point being [1:16:10] that this is, I think, [1:16:11] where the anger comes from. [1:16:13] Why would a company [1:16:14] that's a multi-trillion dollar company [1:16:16] need any money? [1:16:17] There was a situation [1:16:18] in Ohio. [1:16:18] Carlisle Group came in, [1:16:20] built a data center in Akron [1:16:21] and we gave them [1:16:22] four and a half million dollars. [1:16:23] Like why? [1:16:24] And I've said [1:16:25] to David Rubenstein [1:16:26] and Carlisle, [1:16:27] give them money back. [1:16:28] You don't need the money. [1:16:29] And if you need the money, [1:16:31] you have a shitty business model. [1:16:33] That's the problem. [1:16:33] So number two, [1:16:35] start with you, [1:16:35] this crappy business model. [1:16:39] See, the best part [1:16:40] of being Catholic, [1:16:41] Sunday, confession, [1:16:42] it's all over. [1:16:44] So switching to power. [1:16:46] President Trump put in place [1:16:48] the Ratepayer Protection Pledge. [1:16:50] It's very simple. [1:16:51] I live in Avon, Ohio. [1:16:53] I don't want to compete [1:16:54] for electricity [1:16:55] with a giant data center. [1:16:56] So data centers [1:16:57] should not add [1:16:58] to the cost of electricity [1:17:00] for average working Americans. [1:17:01] Would you agree? [1:17:03] I would agree. [1:17:04] Go down. [1:17:05] Yes. [1:17:05] Yes. [1:17:06] Of course. [1:17:07] Yes. [1:17:08] Yes. [1:17:08] Right. [1:17:08] So you represent the industry. [1:17:10] So you can, [1:17:11] with full confidence, [1:17:12] say there's no scenario [1:17:13] on earth [1:17:14] in which a data center [1:17:15] is built [1:17:16] that doesn't either [1:17:17] generate its own power [1:17:18] or does not have [1:17:20] an increased impact [1:17:22] on electricity prices. [1:17:23] You would advocate [1:17:24] to say you should not do that. [1:17:25] Senator, speaking in general, [1:17:28] as I can't speak [1:17:29] for every data center [1:17:30] in the country, [1:17:31] but certainly ITI [1:17:32] is supportive [1:17:33] of more electrons [1:17:34] than the grid bringer [1:17:34] on generation [1:17:35] getting data centers [1:17:36] the power they need. [1:17:37] But no impact [1:17:38] on electricity. [1:17:39] That's really important. [1:17:41] And I applaud [1:17:41] President Trump [1:17:42] for that Ratepayer Protection Pledge. [1:17:43] Number three, [1:17:45] environmental impact. [1:17:46] These systems need [1:17:47] to be closed loop [1:17:47] and have no impact [1:17:49] on the environment [1:17:50] or a positive impact. [1:17:51] Agreed? [1:17:53] Agreed. [1:17:54] Yes. [1:17:54] Yes, very much so. [1:17:57] Sounds right. [1:17:58] As a generality, yeah. [1:17:59] Right, because we [1:18:00] screwed that up in Ohio. [1:18:01] The Ohio EPA [1:18:02] put out a statement [1:18:03] that said we were [1:18:03] going to lower [1:18:04] the water runoff [1:18:06] requirements for data centers. [1:18:07] That's next level insanity. [1:18:09] The last one, [1:18:10] when these data centers [1:18:12] come in, [1:18:12] they should have [1:18:13] a positive impact [1:18:14] in the community [1:18:14] in some way. [1:18:16] What that means [1:18:17] is up to that local community, [1:18:18] whether it's [1:18:19] workforce grants, [1:18:20] improving infrastructure, [1:18:21] but they need [1:18:22] to be good public citizens. [1:18:24] Would you agree with that? [1:18:26] I would agree with that [1:18:27] and I think that the community [1:18:28] should be the leader. [1:18:29] Yeah, absolutely. [1:18:29] Whatever that community needs. [1:18:32] Yes, obviously. [1:18:33] And these data centers [1:18:34] do bring in lots of money [1:18:35] in property taxes. [1:18:36] So I think that there is... [1:18:37] But so does every single [1:18:38] solitary commercial business. [1:18:39] Of course, yes, of course. [1:18:40] Right. [1:18:41] Yes, sir. [1:18:42] Yes, sir. [1:18:43] Okay, so these are... [1:18:44] I think if we can get [1:18:45] through these four principles [1:18:46] and data centers [1:18:47] behave this way, [1:18:48] the operators, [1:18:49] the hyperscalers [1:18:49] behave this way, [1:18:50] I think we could get [1:18:51] this conversation [1:18:52] in a different place. [1:18:53] And I think this is [1:18:54] very, very common sense. [1:18:56] Real quickly, [1:18:57] Mr. Fife, [1:18:58] I'll ask you the question. [1:18:59] Shifting gears completely. [1:19:02] We have vehicles [1:19:03] that are made by China [1:19:04] that connect [1:19:06] into critical infrastructure, [1:19:08] record videos, [1:19:09] send back data, [1:19:10] send back high-definition photos. [1:19:14] On a scale of really, [1:19:15] really, really crazy stupid [1:19:16] to common sense, [1:19:18] where would it be [1:19:19] to allow these cars [1:19:20] to come into the United States? [1:19:21] All right. [1:19:22] Full-on crazy stupid. [1:19:23] Catastrophically stupid. [1:19:24] Okay, great. [1:19:25] Thank you. [1:19:25] You gave time back. [1:19:34] That's fantastic. [1:19:37] All right, Senator Smith. [1:19:38] Well, thank you, [1:19:38] Mr. Chair, [1:19:39] Ranking Member. [1:19:40] Welcome, everybody, [1:19:40] to the committee. [1:19:42] Dr. West, [1:19:43] I'd like to start with you. [1:19:44] I'm very interested [1:19:45] in the intersection [1:19:46] between AI and housing, [1:19:48] and fair housing in particular, [1:19:49] and I want to talk with you [1:19:50] about that for a little bit. [1:19:51] So, you know, [1:19:52] Congress passed [1:19:52] the Fair Housing Act [1:19:53] in, I believe, 1967 or 68. [1:19:56] Actually, Walter Mondale, [1:19:57] my home state senator, [1:19:59] led the way on that. [1:20:00] And it's been a long-standing [1:20:02] tenant of our country [1:20:03] that we should not be discriminating [1:20:04] when it comes to housing. [1:20:05] And so, interesting [1:20:07] how AI intersects [1:20:08] with this, right? [1:20:09] Because we know [1:20:11] that AI tools can, [1:20:13] you know, [1:20:13] there's ways [1:20:13] that they could be [1:20:14] beneficial to housing. [1:20:15] They could make it easier [1:20:16] to build. [1:20:16] They could make it more affordable. [1:20:17] They could make it more accessible. [1:20:19] But also, experts have warned [1:20:20] that without [1:20:22] anti-discrimination protections, [1:20:24] they could also, [1:20:25] you know, create risks [1:20:27] for renters and homeowners. [1:20:28] We've all heard about [1:20:29] this sort of, you know, [1:20:30] the algorithmic pricing of rent [1:20:32] as just one example of that. [1:20:34] So, this is just [1:20:34] an open-ended question to you. [1:20:36] Would you, you know, [1:20:37] could you talk a little bit [1:20:38] about this [1:20:39] and what you see [1:20:40] as the opportunities [1:20:40] as well as the challenges [1:20:42] of AI when it comes to housing [1:20:45] and particularly [1:20:45] fighting housing discrimination? [1:20:50] Yeah, I'm so glad [1:20:51] that you brought this [1:20:51] into the conversation [1:20:52] because it's one of the things [1:20:53] that the public sort of directly [1:20:55] is interfacing with AI systems, [1:20:57] but the kind of AI [1:20:58] that is more used on us [1:21:00] rather than by us, right? [1:21:01] It's AI that's used [1:21:02] to set the price of rent [1:21:04] or to determine [1:21:04] what the cost [1:21:05] of somebody's mortgage [1:21:06] is going to be. [1:21:08] And in those uses, [1:21:09] we see a few particular effects. [1:21:12] One is that [1:21:13] these systems have a tendency [1:21:15] to sort of scale [1:21:18] and amplify existing patterns [1:21:20] of inequality across society. [1:21:22] But in doing that, [1:21:23] they also make it harder [1:21:24] for the average American [1:21:26] to see the effects on, [1:21:28] it's harder to see [1:21:29] that you are being discriminated against [1:21:31] and to be able to take action. [1:21:33] It's sort of like [1:21:34] you don't know [1:21:34] what you don't know scenario [1:21:35] when it comes to how your, [1:21:37] what rent you're being charged [1:21:38] or what your mortgage rate is, [1:21:39] for example. [1:21:40] Exactly. [1:21:40] And it ratchets up [1:21:41] the information asymmetries [1:21:43] between everyday people [1:21:45] and people who are in positions [1:21:46] of relative power. [1:21:48] Landlords, [1:21:50] you know, [1:21:50] mortgage lenders [1:21:51] sort of have greater [1:21:53] insight and information [1:21:55] and access to data [1:21:57] that they can use [1:21:58] in ways that, [1:22:00] you know, [1:22:00] intentionally or unintentionally [1:22:02] have the effect [1:22:03] of ratcheting up inequality [1:22:05] that everyday people are feeling. [1:22:07] And so we have, [1:22:08] and we have laws [1:22:09] that say that you can't discriminate [1:22:10] against somebody [1:22:11] when it comes to housing. [1:22:12] But if we're not enforcing those laws, [1:22:14] which I think some of us [1:22:14] have real concerns about [1:22:15] with this administration, [1:22:17] then doesn't that then amplify [1:22:19] and amplify the discriminatory impact [1:22:22] of some of these AI tools [1:22:24] and also just make it [1:22:25] so much harder for people [1:22:26] to get any, [1:22:27] you know, relief? [1:22:31] I think that's exactly right. [1:22:33] It's both the, you know, [1:22:35] lack of enforcement compounds [1:22:36] with the inability of individuals [1:22:39] to then go and seek remedy. [1:22:41] So this is something I think [1:22:42] that we need to really pay attention to, [1:22:43] Mr. Chair, [1:22:44] as we think about AI impacts. [1:22:46] I want to just touch [1:22:47] on one other thing. [1:22:48] There have been some very interesting proposals [1:22:49] that have come out [1:22:51] in the last weeks, I'll say, [1:22:52] around how AI systems [1:22:57] that are trained [1:22:58] on the collective, you know, [1:23:03] knowledge and data of Americans, [1:23:06] that the wealth that is generated [1:23:08] by those systems should be shared, [1:23:09] whether it's in a sovereign wealth fund. [1:23:11] Senator Sanders has a proposal. [1:23:13] I believe even Sam Altman [1:23:15] has had a proposal. [1:23:16] And so could you just talk briefly [1:23:18] about what you think [1:23:20] some of these options are [1:23:21] and how we ought to think [1:23:22] about as policymakers how people [1:23:25] that are generating the data [1:23:27] and the information [1:23:28] that is fueling this incredible boom [1:23:30] in wealth for very few people [1:23:32] could actually be shared [1:23:33] by everybody? [1:23:36] Sure. [1:23:37] I mean, it's good to see [1:23:38] an interest in approaches [1:23:40] that cut right to the heart [1:23:41] of AI companies' political [1:23:43] and economic power. [1:23:44] I think that that's the right place [1:23:46] to be starting from. [1:23:48] I think it's equally important [1:23:49] that we get the incentive structures right. [1:23:52] And so I think that there's [1:23:53] two components here. [1:23:55] One is that we make sure [1:23:57] that there's an approach [1:23:58] and policy that provides [1:24:00] for the kind of safety net needed [1:24:01] that's to, you know, [1:24:03] help the public to whether [1:24:05] whatever changes might be coming. [1:24:08] But equally that we also ensure [1:24:10] that our regulators have the right [1:24:13] kind of accountability measures [1:24:14] in place to make sure [1:24:15] that the actual use of these tools [1:24:17] isn't then going to cause [1:24:18] other kinds of harm. [1:24:19] Right. [1:24:19] So just to be blunt about it, [1:24:21] if you're creating [1:24:22] an incentive structure [1:24:23] where the regulatory, [1:24:24] the public sector regulatory function [1:24:26] is also, on the other hand, [1:24:28] benefiting from the wealth [1:24:29] that's being generated [1:24:30] by the entity being regulated, [1:24:32] that could potentially [1:24:33] be a negative incentive. [1:24:34] I think that's the tension [1:24:36] that we need to resolve [1:24:37] in the policy approach here. [1:24:38] Yeah. [1:24:38] Okay. [1:24:39] Thank you very much. [1:24:41] Thank you. [1:24:42] Senator Banks. [1:24:43] Thank you, Mr. Chairman, [1:24:44] first of all, [1:24:45] for holding this hearing. [1:24:47] There is no contest [1:24:48] that matters more [1:24:49] to America's future [1:24:50] than the race to beat China [1:24:51] in artificial intelligence. [1:24:53] And I share your commitment [1:24:55] to this issue. [1:24:55] Really appreciate you [1:24:56] organizing this hearing. [1:24:58] The nation that leads in AI [1:25:00] will set the terms [1:25:01] of the world's economy [1:25:02] and command the high ground [1:25:04] in military power. [1:25:06] That's why I believe [1:25:07] that the United States [1:25:08] must lead, [1:25:08] and so does President Trump. [1:25:10] America has to make [1:25:11] the best chips, [1:25:13] develop the best technology, [1:25:15] and protect what we build. [1:25:17] Our most advanced AI chips [1:25:19] are the crown jewels [1:25:20] of American power, [1:25:22] and we cannot let them [1:25:23] end up in the hands [1:25:24] of the Chinese Communist Party. [1:25:26] Our most cutting-edge AI chips [1:25:28] have been smuggled [1:25:29] to Chinese military companies [1:25:31] through third nations, [1:25:33] and China is using these chips [1:25:35] to build AI models [1:25:37] that rival ours. [1:25:38] Strong export controls [1:25:40] will keep America ahead [1:25:42] instead of handing [1:25:42] the lead to Beijing. [1:25:44] I have worked on legislation, [1:25:47] including the AI Overwatch Act [1:25:49] with Congressman Brian Mast, [1:25:51] the chairman of the House [1:25:53] Foreign Affairs Committee, [1:25:54] to make sure that [1:25:55] the Chinese military [1:25:56] cannot access [1:25:57] any of our highest-end chips, [1:25:59] just as President Trump has said. [1:26:01] Mr. Feith, [1:26:02] my bill, the AI Overwatch Act, [1:26:04] is designed to put [1:26:05] America's export control enforcement [1:26:07] on a firm footing. [1:26:09] It would stop [1:26:10] the most capable chips [1:26:11] from going to Beijing [1:26:11] and building the Chinese military. [1:26:13] When President Trump [1:26:15] sits across from Xi Jinping, [1:26:17] it strengthens his hand [1:26:19] to say that Congress [1:26:20] has his back [1:26:21] and that our most advanced chips [1:26:23] are not for sale [1:26:24] to our adversaries. [1:26:25] You already responded [1:26:26] to ranking member Warren [1:26:28] earlier that you support [1:26:30] the AI Overwatch Act, [1:26:32] but can you talk more [1:26:33] about how ensuring [1:26:34] that China doesn't gain access [1:26:36] to leading-edge chips [1:26:37] protects America's [1:26:39] national security [1:26:40] like what the AI Overwatch Act [1:26:42] is designed to do? [1:26:43] Yes, thanks, Senator, [1:26:45] and thanks for leading on that, [1:26:46] working with Chairman Maas [1:26:49] in the House [1:26:49] and others, [1:26:50] including your Democratic [1:26:51] colleagues on it. [1:26:52] The importance, [1:26:54] as you said, [1:26:55] is partly to codify [1:26:57] some of the Trump administration's [1:27:00] own policy [1:27:00] that we're not going to send [1:27:01] our leading-edge chips [1:27:02] to China, [1:27:03] and then putting in [1:27:04] a structure [1:27:05] for, I think, [1:27:07] a really important enhancement [1:27:08] of the relationship [1:27:09] between the commerce [1:27:10] and, rather, [1:27:11] the Congress [1:27:11] and the executive branch [1:27:12] on how we do exports [1:27:14] of these most critical technologies [1:27:15] because, like you said, [1:27:16] they're fundamental [1:27:17] for military purposes, [1:27:19] for all sorts of security [1:27:20] and intelligence [1:27:21] and human rights purposes, [1:27:22] and they're fundamental [1:27:23] to the strategic competition [1:27:25] broadly [1:27:26] between the U.S. [1:27:26] and China, [1:27:27] our tech companies [1:27:28] and theirs [1:27:29] all around the world. [1:27:31] And some of the critics [1:27:33] of AI Overwatch [1:27:33] suggest that somehow, [1:27:35] I mean, [1:27:35] this seems ridiculous to me, [1:27:38] but somehow [1:27:38] that AI Overwatch [1:27:40] or writing export controls [1:27:42] into the law [1:27:43] might weaken our hand. [1:27:45] I mean, [1:27:45] that's outrageous, [1:27:46] isn't it? [1:27:47] I think that sounds [1:27:48] right to me. [1:27:50] You have also made [1:27:51] the point today [1:27:52] that the United States [1:27:53] writes strong export [1:27:55] control rules [1:27:56] that they often [1:27:57] fail to enforce, [1:27:58] and we continue [1:28:00] to see smuggling networks, [1:28:02] moving billions of dollars [1:28:03] in high-end chips [1:28:04] to the Chinese military, [1:28:06] much of it routed [1:28:06] through Southeast Asia. [1:28:08] There are other [1:28:09] export control bills [1:28:10] like Senator Cotton's [1:28:12] Chip Security Act. [1:28:13] Senator Ricketts [1:28:14] has been a great leader [1:28:15] on this issue [1:28:15] through the Match Act, [1:28:17] both bills [1:28:18] that I fully support [1:28:19] and I've co-sponsored. [1:28:21] There are common sense policies [1:28:22] that address national security [1:28:23] while also ensuring [1:28:25] American economic competitiveness [1:28:27] is not harmed. [1:28:28] What is the biggest [1:28:29] gap in our export control [1:28:30] enforcement today [1:28:31] and how can this committee [1:28:33] seek to fix it? [1:28:35] Well, I think these bills [1:28:36] are especially important. [1:28:37] I mean, bills that [1:28:38] focus on smuggling, [1:28:41] focus on the ability [1:28:41] to enhance things [1:28:44] like location verification [1:28:45] on chips. [1:28:47] You know, [1:28:47] just in the last 13, [1:28:50] 14 months or so, [1:28:51] we've had some influential voices, [1:28:54] you know, argue that [1:28:55] essentially there's no such thing [1:28:57] as smuggling. [1:28:58] It might be physically impossible [1:28:59] was the claim. [1:29:01] Now we've seen mounting evidence, [1:29:03] evidence in news reporting, [1:29:06] evidence in the indictment [1:29:08] by the Southern District [1:29:09] of New York recently [1:29:10] of some leadership [1:29:13] of one of the biggest [1:29:13] server companies [1:29:14] in the world. [1:29:15] And there's a sense [1:29:18] that there's a lot more out there, [1:29:19] I think, [1:29:20] in the intelligence record [1:29:21] and in the sort of things [1:29:23] that our export control [1:29:25] bureaucracy is able to see [1:29:27] but hasn't always been able [1:29:30] to take full action on [1:29:32] or prevent, [1:29:33] and I think legislation [1:29:34] that focuses on it [1:29:35] as well as budget support. [1:29:38] But I think this is fundamentally [1:29:39] a matter of real policy [1:29:41] prioritization [1:29:41] and then the budget [1:29:43] would follow that. [1:29:44] Thank you. [1:29:44] My time has expired. [1:29:47] Senator Kim. [1:29:48] Yeah, thank you. [1:29:49] Thank you all for joining up. [1:29:50] I just have to say, [1:29:52] I'm so glad [1:29:53] that the chairman [1:29:54] called for this hearing. [1:29:55] I think this is one [1:29:55] of the most consequential topics [1:29:57] that I've had to engage on [1:29:59] in a hearing [1:30:00] in my time [1:30:01] in the Senate so far. [1:30:02] And I think that, [1:30:03] from what I've heard, [1:30:04] there's wide agreement [1:30:05] amongst my colleagues [1:30:06] about the importance [1:30:08] for export controls [1:30:09] and for engaging. [1:30:10] I thank Senator Ricketts [1:30:11] for working with me [1:30:12] on the Match Act, [1:30:13] Senator Banks, [1:30:14] and others [1:30:14] that I've had a chance [1:30:15] to work on. [1:30:16] But I think what I'm trying [1:30:17] to hit home here, [1:30:18] Mr. Feith, [1:30:18] I'd like to just connect [1:30:20] with you on this [1:30:20] because, you know, [1:30:21] certainly the administration [1:30:23] has taken some actions, [1:30:24] but how important is it [1:30:26] that we push forward [1:30:28] on this legislation [1:30:28] and try to create [1:30:30] an enduring [1:30:30] and comprehensive framework? [1:30:32] How important is that [1:30:33] for us to be able [1:30:34] to take up here in Congress? [1:30:35] I think very important. [1:30:37] I mean, [1:30:37] there's a really important role [1:30:38] for statute [1:30:39] that lasts. [1:30:40] Because I would say [1:30:41] that the temporary nature, [1:30:43] if the idea is [1:30:44] that these export controls [1:30:45] could very well shift [1:30:46] from administration [1:30:47] to administration, [1:30:49] how does that affect [1:30:50] how allies might engage [1:30:52] or how our adversaries [1:30:53] might engage [1:30:54] or how business might engage? [1:30:56] I think it's hard [1:30:58] to generalize. [1:30:59] I mean, [1:30:59] we've got qualities [1:31:00] of our political system [1:31:01] that are designed [1:31:01] for steadiness [1:31:04] and for, you know, [1:31:05] long-term codification, [1:31:06] and we've got aspects [1:31:07] where the discretion [1:31:08] of an executive [1:31:08] is important [1:31:09] and that's part of why [1:31:10] we have cycles [1:31:11] in our politics. [1:31:12] So, you know, [1:31:13] I think some of that [1:31:14] is natural [1:31:14] and appropriate, [1:31:15] but I think [1:31:16] to your fundamental point, [1:31:18] Congress has [1:31:19] an enormous role [1:31:20] in this part [1:31:23] of this national security challenge. [1:31:26] This committee, [1:31:27] of course, [1:31:27] has oversight [1:31:27] of export controls, [1:31:29] and there are [1:31:31] major opportunities [1:31:31] that I think, you know, [1:31:33] sponsors on this committee [1:31:34] and elsewhere [1:31:34] in the Congress [1:31:35] have pointed out [1:31:35] to really bolster [1:31:36] national security, [1:31:38] strengthen our export control rules [1:31:40] and enforcement capability, [1:31:41] and kind of assert [1:31:42] the Congress's role [1:31:43] in ways that would be lasting [1:31:45] on, you know, [1:31:46] this sort of nicely interlocking [1:31:47] set of aspects [1:31:48] of this national security problem. [1:31:50] And I hope [1:31:51] that we can work together [1:31:51] to build that long-term resiliency [1:31:53] and that direction [1:31:54] for our country. [1:31:56] And look, [1:31:57] given the speed [1:31:58] in which the technology [1:31:59] is advancing, [1:32:00] as well as sometimes [1:32:00] the lag in which it takes [1:32:02] to implement law [1:32:03] and rulemaking, [1:32:04] would you say [1:32:05] that it's important [1:32:06] for us to move [1:32:07] as soon as possible [1:32:07] on this? [1:32:08] Yes, sir. [1:32:09] And when we're talking [1:32:10] about aligning [1:32:10] with our allies, [1:32:11] you know, [1:32:11] we certainly talked about [1:32:12] in terms of [1:32:13] what's going to have [1:32:14] the strongest impact [1:32:15] to impede [1:32:16] and stop gaps [1:32:17] from forming [1:32:18] that our adversaries [1:32:19] can exploit. [1:32:20] But I would also say [1:32:21] that it probably also [1:32:23] has a benefit [1:32:24] of trying to level [1:32:25] the playing field [1:32:26] for U.S. companies [1:32:27] in terms of how [1:32:28] they're engaging [1:32:29] abroad as well. [1:32:30] Yes, [1:32:30] and the level playing field [1:32:32] is especially important [1:32:32] when it comes [1:32:33] to the question [1:32:34] of so-called [1:32:35] allied backfilling. [1:32:37] You know, [1:32:37] cooperation with our allies [1:32:38] is really very important, [1:32:40] including so that [1:32:41] we don't have companies [1:32:41] from important allied countries [1:32:43] that aren't subject [1:32:45] to the same types [1:32:46] of national security restrictions [1:32:47] that U.S. companies [1:32:48] are subject to [1:32:49] because these are [1:32:50] national security concerns [1:32:51] vis-a-vis China [1:32:52] that we share together. [1:32:54] Chair, [1:32:55] I just wanted to say [1:32:56] that there's been [1:32:57] strong support [1:32:58] for the Match Act, [1:32:59] for these different efforts [1:33:00] we're doing [1:33:01] from outside experts [1:33:02] industry stakeholders, [1:33:04] national security experts, [1:33:05] and so to that end, [1:33:07] I ask for unanimous consent [1:33:08] to place a letter [1:33:09] of support [1:33:10] for the Match Act [1:33:10] signed by a broad-based [1:33:12] coalition of stakeholders [1:33:13] into the record. [1:33:14] Without objection. [1:33:16] Mr. Flynn, [1:33:18] I don't necessarily [1:33:19] have a question here, [1:33:19] but I just wanted [1:33:20] to thank you [1:33:20] for bringing up [1:33:21] the importance [1:33:22] of supporting BIS [1:33:24] and making sure [1:33:25] that as we're pushing [1:33:26] forward on the HIS [1:33:26] expert controls [1:33:27] that we have the effort [1:33:28] to be able [1:33:29] to come behind that. [1:33:30] And to my colleagues [1:33:31] here in the committee, [1:33:32] you know, [1:33:33] I'm leading the appropriations [1:33:34] effort to try [1:33:35] to increase funding [1:33:35] to BIS. [1:33:36] I hope that that's something [1:33:38] that we can continue [1:33:38] to push on together. [1:33:40] Dr. West, [1:33:41] I wanted to just talk [1:33:42] to you. [1:33:43] One thing that I hear a lot [1:33:44] of concerns about, [1:33:46] you know, especially [1:33:46] from seniors [1:33:47] and others is about fraud [1:33:49] and what might happen next [1:33:51] when it comes [1:33:51] to AI's capacity [1:33:53] to just amplify [1:33:55] and magnify that [1:33:56] so far beyond anything [1:33:57] we've seen before. [1:33:58] What are the practical steps [1:33:59] that we should be looking [1:34:00] at about fraud [1:34:02] from a government [1:34:03] and industry [1:34:04] and financial institution [1:34:05] section? [1:34:07] Yeah, I'm glad [1:34:07] that you brought this up [1:34:08] because we are seeing [1:34:09] a pattern where scams [1:34:10] and fraud are on the rise, [1:34:12] particularly with the [1:34:13] off-the-shelf [1:34:13] and inexpensive availability [1:34:15] of generative AI tools. [1:34:17] So there's an ecosystem [1:34:19] around fraud [1:34:20] where law enforcement [1:34:21] and social platforms [1:34:22] and financial institutions [1:34:24] and tech companies [1:34:24] all play a part. [1:34:25] I think that we need [1:34:28] to place our greatest emphasis [1:34:29] on the platforms [1:34:31] and financial institutions [1:34:32] that have the greatest insights [1:34:35] through the amount of data [1:34:36] that they're able to collect [1:34:37] to identify scams [1:34:38] and to stop them sort [1:34:41] of means that we need [1:34:42] to be like protecting [1:34:43] against them occurring [1:34:44] in the first place. [1:34:45] So we should be requiring [1:34:46] more of those institutions [1:34:48] as the starting place. [1:34:50] Yeah, thank you so much. [1:34:51] And I just want to end [1:34:51] by just again reiterating [1:34:53] some of my colleagues [1:34:54] on the other side [1:34:54] of the aisle. [1:34:55] I hope we can move forward [1:34:56] on a markup [1:34:57] on the Match Act [1:34:58] and other important pieces [1:34:59] of legislation [1:35:00] in this committee. [1:35:01] And with that, [1:35:01] I'll yield back. [1:35:03] On behalf of the chairman, [1:35:04] Senator Blount Rochester. [1:35:08] Thank you, [1:35:09] Ranking Member Warren. [1:35:11] I have to say, [1:35:12] I literally have so many notes [1:35:14] on my dais here [1:35:15] that I don't even know [1:35:16] where to begin [1:35:16] because this issue [1:35:18] is just so consequential [1:35:19] to our country. [1:35:20] So thank you so much [1:35:21] for your thoughtful testimony [1:35:22] as well. [1:35:23] It's consequential [1:35:24] to our economy, [1:35:25] to our health, [1:35:26] to our national security. [1:35:29] And over the past year, [1:35:31] I asked Chairman Powell [1:35:35] when he was here, [1:35:36] I asked Chairman Warsh [1:35:37] about the impact [1:35:39] that AI is going to have [1:35:41] on our workforce. [1:35:42] And it was interesting to me [1:35:43] because what literally came out [1:35:45] of it was just the uncertainty. [1:35:48] Mr. Reinhart, [1:35:48] you even mentioned data [1:35:49] and the data that we collect [1:35:51] or don't collect. [1:35:52] We had the BLS direct nominee [1:35:54] before us yesterday [1:35:55] and another committee. [1:35:57] And from my time in the House [1:35:58] to being here in the Senate, [1:36:00] in the House, [1:36:01] I started a bipartisan [1:36:01] future work caucus. [1:36:03] And that was almost a decade ago. [1:36:05] And now it is literally here [1:36:06] before us. [1:36:08] And I can say, [1:36:09] I get to look at it [1:36:10] from that lens. [1:36:11] And also, [1:36:11] I've got bipartisan legislation [1:36:13] right now [1:36:14] with Senator Sheehy, [1:36:15] the AI Fraud Accountability Act. [1:36:17] I've looked at it [1:36:18] from AI literacy [1:36:19] to help prepare our families. [1:36:21] And Dr. West, [1:36:23] I think you put it [1:36:24] so perfectly [1:36:25] that we need to put [1:36:26] the needs of the public [1:36:28] at the center of AI policy, [1:36:31] the needs of the public. [1:36:32] And so I want to start off [1:36:34] by saying, [1:36:36] as you said, [1:36:36] there's a lot of feelings [1:36:37] out here about this. [1:36:38] People are either really, [1:36:39] really excited [1:36:40] or really, really scared. [1:36:43] They're excited [1:36:44] about what they maybe can do [1:36:46] from shopping [1:36:47] or decorating [1:36:48] an office, maybe. [1:36:49] Mine. [1:36:50] Or... [1:36:51] Not so much, [1:36:52] but they're going to exist. [1:36:53] Mine. [1:36:53] But they're also afraid [1:36:55] that they're going [1:36:56] to lose their jobs [1:36:56] and that this will have [1:36:59] a detrimental impact. [1:37:00] And so many news reports [1:37:02] recently have come out [1:37:03] about new entrants [1:37:04] into the workforce, [1:37:06] whether it's college students, [1:37:07] women in particular, [1:37:09] and people of color [1:37:10] disproportionately being impacted [1:37:11] by or exposed [1:37:13] to occupations [1:37:14] most vulnerable to AI. [1:37:16] So I'm concerned [1:37:18] that the AI [1:37:18] is moving much faster [1:37:19] than we are [1:37:20] and we're not taking advantage [1:37:22] of the disruptions [1:37:23] or the opportunities. [1:37:24] So Dr. West [1:37:25] and Mr. Reinhardt, [1:37:26] if you could pick [1:37:27] one or two workforce indicators [1:37:30] that Congress [1:37:31] should be closely watching, [1:37:32] what would they be? [1:37:33] I'll start with... [1:37:35] I'll start with you, Dr. West. [1:37:39] Sure. [1:37:39] So obviously, [1:37:40] we want to be looking [1:37:41] at indicators [1:37:42] like layoffs [1:37:43] as a frontline measure [1:37:45] and trying to understand [1:37:46] how that's maybe being [1:37:47] disproportionately felt [1:37:48] across the economy. [1:37:50] I'd also, though, [1:37:50] be looking at things [1:37:52] like employment dislocation. [1:37:54] We've seen, for example, [1:37:56] a number of jobs [1:37:58] being restructured [1:37:59] so that they're sort of [1:38:01] flexible or self-employed [1:38:02] and that also means [1:38:03] that those workers [1:38:04] don't have access [1:38:05] to the same kinds [1:38:07] of social safety net measures [1:38:09] and they're also sort [1:38:10] of dually feeling [1:38:11] the tax burden. [1:38:12] So that's another place [1:38:13] that I'd be looking closely. [1:38:14] Thank you. [1:38:15] Mr. Reinhardt. [1:38:16] Yeah, I would be focusing [1:38:17] on the unemployment rate [1:38:18] within the AI-affected industries. [1:38:20] There has been [1:38:20] a whole bunch [1:38:21] of these indices [1:38:22] that have tried [1:38:23] to understand job loss, [1:38:24] potential job loss [1:38:25] and dislocation. [1:38:27] And I will say [1:38:28] that right now [1:38:28] we're really only seeing it [1:38:30] with entry-level workers, [1:38:32] people that are in the 20-24 range, [1:38:34] 20-25 range, rather. [1:38:36] And I think there's actually [1:38:36] a very good reason for that. [1:38:38] If you're young, [1:38:39] you don't have a lot [1:38:40] of experience, [1:38:40] you don't have skills, [1:38:41] you can't show to employers [1:38:42] that you're skilled [1:38:43] or that you have [1:38:45] the kinds of things [1:38:46] for a job. [1:38:47] And so my sense [1:38:48] is what's happening right now [1:38:49] is that the market [1:38:50] is becoming very noisy. [1:38:52] We talk about this [1:38:52] as a problem [1:38:54] in matching [1:38:54] between employer [1:38:55] and employee. [1:38:57] And so those are [1:38:58] the sorts of things [1:38:59] that I'd be looking towards [1:39:00] is really the unemployment [1:39:01] in the AI-affected industry [1:39:02] and kind of the noisiness [1:39:03] of the market. [1:39:05] I will follow up [1:39:06] with the two of you [1:39:07] and the other panelists [1:39:10] on the data [1:39:11] that we need [1:39:12] and that's required. [1:39:14] We've heard a lot [1:39:15] of promise [1:39:15] that AI will lower cost [1:39:17] and improve affordability. [1:39:19] Dr. West, [1:39:19] what factors [1:39:20] would demonstrate [1:39:21] that those benefits [1:39:22] are actually reaching [1:39:23] consumers and families? [1:39:28] That's a good question. [1:39:29] I mean, [1:39:29] I think the things [1:39:30] that we would be [1:39:31] wanting to look for [1:39:33] are, you know, [1:39:34] general inflationary indicators. [1:39:36] I mean, [1:39:37] honestly, [1:39:38] what we're seeing [1:39:39] is in many sectors [1:39:40] of the economy, [1:39:41] costs are really going up [1:39:42] and this is part [1:39:43] of a generalized trend [1:39:44] toward inflation. [1:39:46] The utility sector, [1:39:47] I think, [1:39:47] is one example of that. [1:39:49] Equally, [1:39:50] we're seeing the rise [1:39:50] of surveillance pricing [1:39:51] where data's being leveraged [1:39:53] against people [1:39:54] in ways that is increasing [1:39:55] prices across the shelf. [1:39:57] Thank you. [1:39:58] I have one more question [1:39:59] here, [1:40:01] and I will send more [1:40:02] for the record. [1:40:02] This is just a yes or no [1:40:03] for everybody on the panel. [1:40:05] Yes or no? [1:40:07] Chairman Scott [1:40:08] talked about [1:40:08] the use of [1:40:09] some of the tools. [1:40:11] We use them every day. [1:40:13] I signed up for one, [1:40:14] loved it, [1:40:15] and within less than a year, [1:40:18] a pricing structure [1:40:19] came out [1:40:20] with multiple tiers. [1:40:24] Are we at risk [1:40:25] of broadening [1:40:26] the digital divide? [1:40:28] Yes or no? [1:40:28] I'm not an expert [1:40:31] on that issue, Senator. [1:40:33] I'm happy to take it back [1:40:34] and give you [1:40:34] some more thoughtful commentary. [1:40:36] Okay. [1:40:36] Mr. Feith. [1:40:38] It's just a yes or no. [1:40:39] Do you think [1:40:40] we're at risk? [1:40:42] I'd have to duck it too, [1:40:44] but I'll... [1:40:45] I appreciate you [1:40:46] saying you ducked it. [1:40:47] Thanks. [1:40:47] Mr. Reif. [1:40:48] Yes, [1:40:49] and I think there are things [1:40:49] that we can be doing [1:40:50] to address it. [1:40:51] Would love to follow up [1:40:52] with you on that. [1:40:53] Yes. [1:40:54] Thank you. [1:40:54] I yield back. [1:40:56] Just in time for [1:40:57] Senator Alston Brooks [1:40:59] to have an opportunity [1:40:59] to share as well. [1:41:01] Thank you so much, [1:41:02] LPR, [1:41:02] for holding the floor [1:41:04] for Senator Alston Brooks. [1:41:05] Thank you so much. [1:41:06] We're so happy you made it. [1:41:07] Thank you. [1:41:08] I'm glad I made it too. [1:41:09] So thank you, [1:41:09] Chair Scott [1:41:10] and Ranking Member Warren [1:41:11] for holding the committee. [1:41:13] Thank you also [1:41:14] to these witnesses [1:41:14] who are here today. [1:41:16] You've heard already, [1:41:17] we've heard the conversation [1:41:19] about artificial intelligence [1:41:21] and the enormous opportunity [1:41:23] that it presents [1:41:24] for our country. [1:41:26] And if we get this right, [1:41:27] I believe that AI [1:41:28] can help American businesses grow, [1:41:30] make government [1:41:31] and financial services [1:41:32] more efficient, [1:41:34] and strengthen [1:41:34] our national security. [1:41:35] And I believe [1:41:36] that the United States [1:41:37] should lead the world [1:41:39] in this technology. [1:41:41] We shouldn't be afraid [1:41:41] of innovation [1:41:42] and we should not create [1:41:44] a regulatory environment [1:41:45] that pushes the next generation [1:41:48] of AI development overseas. [1:41:50] But we also have to remember [1:41:51] that innovation [1:41:52] is most successful [1:41:53] when people have confidence in it. [1:41:56] For our American families, [1:41:57] workers, [1:41:58] small businesses, [1:41:59] and communities, [1:42:00] AI cannot just be happening [1:42:02] around them [1:42:03] or to them, [1:42:04] it has to expand [1:42:05] opportunity for them. [1:42:07] So that means making sure [1:42:08] that people can trust [1:42:09] the systems [1:42:10] that they are using, [1:42:11] whether they're small businesses [1:42:12] who are trying to compete, [1:42:14] a worker learning new tools [1:42:16] or a family applying for a loan, [1:42:18] buying a home, [1:42:19] or managing their finances. [1:42:21] It also means being clear-eyed [1:42:22] about how this technology [1:42:24] is deployed, [1:42:25] especially in areas [1:42:26] where the stakes are high [1:42:27] for consumers, workers, [1:42:28] and the economy. [1:42:28] And we also have to be honest [1:42:30] that for many workers, [1:42:32] this moment brings real questions [1:42:34] about job security [1:42:35] and displacement. [1:42:37] As this technology changes [1:42:38] the way that Americans work, [1:42:40] we have to make sure [1:42:40] that workers are not being [1:42:41] left behind or pushed aside, [1:42:43] but supported with the skills [1:42:45] and training [1:42:45] that can allow them [1:42:47] to use AI to their advantage. [1:42:48] So the question for me [1:42:49] is not whether we should support [1:42:51] AI innovation. [1:42:53] We should. [1:42:54] The question is how we make sure [1:42:55] American innovation [1:42:57] remains fast, [1:42:58] competitive, safe, secure, [1:43:00] and broadly beneficial. [1:43:01] So the first question [1:43:02] I have for Mr. Flynn, [1:43:05] American advantage in AI [1:43:06] depends in part [1:43:07] on whether companies [1:43:08] can build, deploy, [1:43:10] and scale tools securely, [1:43:12] especially in sectors [1:43:13] like finance, [1:43:14] where trust and reliability [1:43:15] are essential. [1:43:16] So when you talk to companies [1:43:17] across the technology sector, [1:43:20] what would you say [1:43:21] they are most worried about? [1:43:23] What are they saying [1:43:24] could slow AI leadership? [1:43:26] Is it cyber vulnerabilities? [1:43:27] Would you say data rules, [1:43:29] supply chain risks, [1:43:30] or something else? [1:43:32] Thank you, Senator, [1:43:33] for that question. [1:43:34] I don't know if there's one thing, [1:43:37] but certainly the supply chain [1:43:38] challenges are very real. [1:43:40] As the demand for AI capabilities [1:43:43] expands, [1:43:45] that does squeeze the supply chain [1:43:48] in a number of areas. [1:43:50] We don't always anticipate [1:43:51] where those market adaptions may be. [1:43:54] So I think it's incumbent upon Congress, [1:43:57] upon all of us [1:43:58] to keep partnering closely [1:44:01] on this challenge [1:44:01] to make sure that the supply chain issues [1:44:03] can be, [1:44:05] do not become a plateau [1:44:06] for our ability [1:44:07] to continue to advance [1:44:08] this technology. [1:44:10] So now in high-stakes sectors [1:44:11] like financial services, [1:44:12] what do you think would help [1:44:14] adopt AI responsibly [1:44:16] while giving consumers confidence [1:44:18] that their sensitive data [1:44:19] is being protected? [1:44:20] The cyber security challenge [1:44:22] is real. [1:44:24] Industry folks were encouraged [1:44:27] by the president's executive order [1:44:29] recently on AI and cyber. [1:44:31] It's a good first start [1:44:33] in terms of creating [1:44:34] the right bureaucratic lanes [1:44:36] and to make sure [1:44:37] that there's good communication [1:44:38] between industry [1:44:39] and particularly affected [1:44:40] critical infrastructure sectors, [1:44:41] including financial services. [1:44:44] But at the end of the day, [1:44:46] in this immediate term, [1:44:48] the information sharing challenge [1:44:49] is going to be very real, [1:44:50] which is why I've talked [1:44:52] so much today [1:44:53] about the need [1:44:53] to reauthorize [1:44:56] the Cyber Information Sharing Act [1:44:57] of 2015, CISA 2015, [1:44:59] that lapses [1:45:00] at the end of this year. [1:45:01] If it lapses, [1:45:02] that would be [1:45:02] an extraordinary unforced error [1:45:04] in terms of the average American's [1:45:06] ability to trust, right, [1:45:09] that their systems [1:45:10] are being well defended. [1:45:11] Now, lastly, [1:45:12] to Dr. West. [1:45:14] So I want to turn [1:45:15] to public trust, [1:45:16] which I see [1:45:16] as an indispensable part [1:45:18] of innovation. [1:45:19] So if people, for example, [1:45:21] can't trust AI systems, [1:45:22] they just simply won't use them. [1:45:24] If small businesses [1:45:25] don't understand the rules, [1:45:26] then they'll hesitate [1:45:26] to adopt them. [1:45:27] And if families are worried [1:45:29] that AI is deployed [1:45:30] in such a way [1:45:31] that they cannot see, [1:45:33] understand, [1:45:33] or challenge, [1:45:34] then we've built a system [1:45:35] that can't be used. [1:45:36] So what would it look like [1:45:38] to build trust [1:45:38] in a way that continues [1:45:40] to support innovation [1:45:41] so that the American people [1:45:42] are able to take advantage [1:45:44] of the opportunities [1:45:45] that these new technologies provide? [1:45:47] I think trust will come [1:45:51] when we have the right kind [1:45:52] of sort of institutional measures [1:45:55] in place [1:45:55] that make sure [1:45:56] that these systems [1:45:57] are being adequately tested [1:45:59] and vetted [1:45:59] before they're, you know, [1:46:01] before they're impacting people [1:46:02] or causing harm. [1:46:04] We don't have, you know, [1:46:06] baseline mechanisms, [1:46:07] especially in, like, [1:46:08] these critical infrastructure areas [1:46:10] that are requiring [1:46:11] the systems to be evaluated [1:46:14] if they're fit for purpose, [1:46:15] if they're efficacy, [1:46:17] their safety, [1:46:18] and I think that, you know, [1:46:19] focusing in there, [1:46:21] making sure that this is happening [1:46:23] before systems are entering [1:46:25] the market is the right place. [1:46:26] Otherwise, we're kind of left [1:46:28] for companies [1:46:28] to be grading their own homework. [1:46:30] Thank you. [1:46:30] Great. [1:46:30] Thank you. [1:46:31] Well, thank you all [1:46:32] for being present with us today. [1:46:34] It's been quite informative, [1:46:35] and we really appreciate [1:46:36] this hearing is adjourned.