Former FTC Technologist Warns Against an AI ‘Cartel’

0:00 We think antitrust, we think monopolistic or anticompetitive practices. The Frontier Labs are saying, Neil, we need this waiver so that we can work together. Your response? So there's so many ways that companies can work together that are not anticompetitive. But, obviously, antitrust has a big role in making sure that when companies collaborate, they're not doing things that ultimately harm of competition or consumers and the results that they expect from a competitive environment. And so I think that there are like I said, there are if you're building safe products and you need to build an industry standard or best practices to do that, there's lots of ways that these companies can collaborate. Most of all, they can try internally, to build their products safely, and that's actually, they're obligated in many ways to do that.

0:53 You are in it seems in a school of thought shared by lots of of leaders in AI that at the end of the day, the Frontier Labs should work this out for themselves if they have command of the technology. Regulatory capture is a concern. It's the idea that those Frontier Labs has helped set the rules that's to their benefit. Your interpretation. So regulatory capture is a huge problem, but what we know from history is that cartels are fragile. The only cartels that are durable are ones that are endorsed by the government.

1:27 And so I think that is the risk that David Sachs is pointing to, that the types of exemptions that they're seeking here could make a durable cartel that would ultimately harm competition and harm The US, ability to stay at the frontier. Are existing laws, particularly those around illegal action, sufficient right now, Neil, to your mind? In other words, if a model does something that constitutes a criminal offense, there is a safety net of the law anyway.

2:00 I I think that's right. Especially in the liability space, our existing law says that, you know, if if you're doing something that these if these companies are training models, they're and they're not doing it safely enough and those models cause harm to third parties, I think the law is pretty clear, especially FTC consumer protection law, and I think many tort laws, that those companies will be responsible for those harms. And I think that is part of what's driving Dario's call for pacing, and it's part of what is making the industry say, hey. We we need to make sure that we're we ourselves are being cautious that we're creating a product and that we're creating it in a way that isn't causing harm to third parties.

2:45 You were formerly the chief technologist at the FTC. Right? Could you take me inside the organization a week like this? How would the FTC be looking at, thinking about, responding to what has been a a roller coaster of events? Buzz, I think for sure, they would be looking at the various authorities that they have both on the competition side and as well on the consumer protection side. And they'd be looking for unfair methods of competition. They'd be looking for unfair or deceptive acts or practices that might be occurring here. And the FTC, in particular, has a long history of data security enforcement under its consumer protection, unfair and and deceptive acts and practices. And I think here, they would be looking for some key indicators that what the companies are doing is not reasonable. It does not match best practices, and that it could be easily avoided by the companies with some extra, precaution. And, they will be looking to see if there's harm that's caused and whether or not, the harms that are caused are outweighed by the benefits to those specific practices that the companies are doing. So I think there's a lot for them to think about, and and they're doing it.

3:55 Neil, the harm that's imagined by some, not all some, is that AI posts, poses an existential threat to human life. Right? That's if even if you go about ten days from the anthropic research of the resigns, I mean, how does the FTC account for that harm, potential harm to a consumer? It's it's a great question. You know, obviously, those types of, speculative harms, one would hope in many ways that those would show up as much smaller types of incidents early on, and that does seem to be the case here. And so so we're seeing incidents, where there is not a lot of harm that's happening, but where it raises some red flags.

4:38 And I think what that means is there's time to look at the laws, to look at the practices in the industry, and say, like, hey. We need to change course here. We need to correct things. How do we do that? Do we need some ex extra laws? We need to see if there's gaps in the law to fill. But I I think we are getting warning signs here that are the exact types of inter that start this iterative process of saying, hey. How can we do better, and how can we avoid these, bigger harms in the future?

5:05 You you're at the Abundance Institute now, an organization who wants technology to grow and thrive. Right? You know, there's a lot of positivity around AI. It's useful. There is some economic upside. Does the FTC ever take that into account? You know, again, drawing on your past experience and your and your current research on, like, how it's good for people. Oh, absolutely. I mean, I think that's a big part of the analysis both on the competition side where the question is, are these practices creating consumer welfare? Are they improving consumer welfare? And then on the consumer protection side, where the question in the unfairness analysis in particular is whether any harms caused by a practice are outweighed by benefits. And I think in the AI space in particular, there are enormous benefits. We're seeing that in stock values.

5:54 We're seeing that investment, and we're seeing it in the huge amount of adoption and usage that individuals are doing. I think we're nowhere near the top of the types of problems that a general purpose AI system can help us with. And so I think there are enormous benefits here. We need to mitigate the harms, but we need to not lose sight of the fact that that mitigation, we wanna make sure it doesn't prevent the the best possible world of the the benefits that could come.