Anthropic’s powerful new AI model raises concerns about high-tech risks

[0:00] ANTHROPIC announced this week it has begun limited testing of its newest AI model called Mythos, [0:05] one the company says is so powerful it could cause widespread disruption if released to the public. [0:11] It's just generally better at pursuing really long-range tasks that are kind of like the tasks [0:18] that a human security researcher would do throughout the course of an entire day. [0:23] Obviously, capabilities in a model like this could do harm if in the wrong hands. [0:28] And so we won't be releasing this model widely. [0:31] For now, ANTHROPIC is giving more than 40 tech companies, including some rivals, [0:36] access to Mythos to test it and identify vulnerabilities across systems. [0:41] But even that move is raising concerns. For a closer look at all of this and the implications, [0:45] we're joined now by Garrett Vandewick, who covers AI for The Washington Post. Thanks for being with us. [0:51] Of course. [0:52] So help us understand the concern here. What specifically makes this model different from [0:56] other AI models and why is there so much, frankly, fear around it? [1:02] The specific concerns that are being called out here is that this model is really good at finding [1:07] gaps in software that hackers could exploit. So right now, all software has bugs, [1:12] but software is pretty complicated and you need to kind of really know what you're doing in order to [1:16] sift through all that code to find something that you could then use to hack into a system. And what [1:21] ANTHROPIC is saying and some of the independent cybersecurity experts that they've also given [1:26] access to this model to are saying is that this can essentially do that automatically. It can sift [1:31] through all sorts of code. Something that might take humans who are very good at this months to do, [1:37] it can do in minutes or hours. And so the concern here is that if this is sort of out in the public, [1:43] anyone can use it, that anyone who wants to hack into any kind of software for whatever reason [1:48] would be able to do it using this technology. And that's why the company is saying at least they're [1:53] sort of keeping it under wraps for now. Keeping it under wraps, but also giving, [1:57] as we mentioned, some 40 other companies, including Microsoft and Nvidia access in part to strengthen [2:03] their own cyber defenses. What do we know about that decision? Does sharing it more widely [2:07] actually reduce the risk or potentially increase it? Yeah, I mean, there is a bit of a precedent here [2:14] in cybersecurity. Often if one company finds some lack in another company's software, instead of [2:20] just giving it to the public and, you know, creating a situation where that other company [2:24] could be hacked, they will sort of go, you know, behind the scenes and say, hey guys, we found this. [2:29] You might want to fix this before the rest of the world figures it out. And so I think it's sort [2:33] of in that tradition that they're doing this. But of course, some people are saying, hey, now we have [2:37] all these powerful tech companies that have access to this allegedly extremely powerful tool for [2:42] cybersecurity. Well, is it also powerful for other things, you know, other things that they could use [2:47] to, you know, increase their business, get an edge on other companies? So there are some complaints [2:52] that, you know, if this thing is really so good, why don't you let the rest of the world actually see [2:55] it for themselves? And then we can decide what to do with it. Logan Graham, who's one of Anthropix [3:01] researchers, suggested that if this AI program were fully released, it could force widespread software [3:08] updates, eventually exposing weaknesses everywhere. Is that a realistic scenario or is he in some ways [3:14] overstating it? Yeah, potentially. I mean, it's difficult because, you know, besides these companies, [3:21] no one has really been able to get their hands on it. I think we always need to take these big AI [3:25] companies with a grain of salt. It's not the first time an AI company has said, oh my goodness, [3:29] our new technology is so powerful, we should be afraid of it. You know, it's great marketing, [3:34] right? Because if something is so powerful that it could, you know, change the world or cause chaos, [3:39] it's also very powerful for doing other things. And so I think we need to be careful. You know, [3:44] I don't, I'm not necessarily saying that Anthropic is lying or misleading the public here. I, I'm sure [3:49] they are very legitimate about these concerns, but I do think that we're already in a situation where [3:53] cybersecurity is pretty atrocious. I mean, everyone's personal data has been hacked at some point. [3:59] If anyone really wants to get into a software system, if they have the resources, [4:03] the, you know, incentive, they will probably be able to do it. We already live in a world where [4:08] software is broken and needs to be updated constantly, right? Every time you open your [4:13] operating system, it's probably pinging you to update the apps that you have on your computer, [4:18] right? That's because of the cybersecurity situation we have right now. And in the same way that this [4:24] mythos technology could be used to hack into computers, it could also be used to defend against [4:30] hacks. And so a lot of the cybersecurity experts are saying, look, yes, this is concerning, but we [4:35] can also use this technology. The good guys can also use it to protect us. And so it doesn't [4:40] necessarily completely change that balance of power that we have right now. Well, say more about that, [4:45] because there is this strange disconnect where you have now even the AI companies themselves warning [4:50] about the potential dangers. And this is as the AI companies are also racing to release more powerful [4:56] systems at the same time. What accounts for that? Yeah. I mean, I think, you know, it's very easy to [5:02] sort of point that and say like, look, like what's really going on here. And I think, you know, each AI [5:07] company is slightly different. They have different incentives, but it's true. I mean, they are all in this [5:11] extremely competitive race to build the best AI system. It's very expensive to train these things. It costs [5:19] hundreds of millions of dollars to develop each new version of this AI technology, and very few [5:25] companies are able to do it. And the entire tech industry is in agreement that this is, you know, [5:29] the most important technology to come out probably since the internet itself. And so there's a huge [5:34] amount of money that is incentivizing the development of this technology. At the same time, a lot of the [5:39] people who work at these companies do legitimately believe that there are concerns that it could be [5:44] used for cybersecurity. It could be used for misinformation. It could, you know, some people even [5:48] believe that it could, you know, become so smart in the coming years that humans are, you know, [5:53] challenged to keep it under control. And so I do think that those are real beliefs held by some [5:58] people at these these these companies. And yet they are locked in this competitive dynamic. [6:03] Garrett DeVink covers AI for The Washington Post. Garrett, thanks again for being with us. [6:19] Support journalism you trust. Support PBS News. Donate now or even better, start a monthly contribution today.