00:00I actually had a whole line of questioning on something, but something Ms. Castor just said,
00:05I may ask all five of you this question, because I think it's fairly important.
00:10You just said there's no indefinite retention on AI, and we're not going to exclude that, correct,
00:17Chairman Khan? Under COPPA, for certain entities in certain contexts. So my question is,
00:24is once information is generated into AI, how is it not retained permanently?
00:32So you mean, basically, how can you require unlearning from the model?
00:35Yeah. Yeah. I think that's a great question.
00:37As far as we know, it's not possible, and that's why timely action is incredibly important,
00:43and making sure firms are not, in the first instance, able to use unlawfully collected data
00:49to train their AI. It also, your point raises another issue, which is business could face an
00:55incentive to actually play fast and loose with the rules, because their monetary gain from that,
01:02and the value they'll generate from training their AI could be so important. And so,
01:06for all of these reasons, stronger rules of the road are going to be critical.
01:10I'm going to ask the two new commissioners one at a time their thoughts on this, because I think
01:14this is the next big conversation. Anybody's heard me talk about this, whether it's government
01:18related, Fourth Amendment related, privately related. If you have a 16-year-old daughter
01:23and a 14-year-old son, privacy in the 21st century is going to be the conversation we have,
01:30and how we monetize all of these things. So what tools do you need from us?
01:38Well, and to your question on what do you do once that data is captured, it's an interesting
01:42question. I'd say, I mean, these emerging technologies, we don't fully understand
01:46exactly what they're capable of, but I think what's interesting is I think there'll be more
01:51technologies that may help us in terms of fighting some of these issues. The FTC's work on the voice
01:57cloning challenge is a great example of where we're concerned about voice cloning, the use of AI,
02:03where you have fraudsters using calling of grandmas and saying, this is your grandson
02:08using a voice clone, and worried about that. So the agency, we just had a voice cloning challenge
02:14where we had companies come in and say, what can we do to fight this? And they provided different
02:19technologies that could help assess that. And I think the same is possible with these types of
02:24issues where you're going to have, we're going to see these types of problems arise in the markets.
02:31I think we'll also see solutions arise too. And I think we can let the markets develop and
02:36understand where those, with constantly assessing where the markets are, remaining vigilant,
02:42using the tools that we have to address the concerns that they have when there are unfair and
02:47deceptive trade practices and, or if there's competition issues as well.
02:51And I know when we talked, coming in, we talked about the Ticket Act, and one of the things that
02:55concerns me about AI is somehow it's the enforcement, right? If you use five other
03:00songs to create a song, you have to pay royalties for it. We all know if you don't, you get sued,
03:06and if you can prove it, you end up losing on a royalty case. There's famous ones. My concern is,
03:11we don't know the five inputs, and there's this fight between open source and non-open source.
03:17And so I don't want to live in a world where every musician is created by AI.
03:22I don't think you get Nirvana in 1991 off of Motley Crue in 1987. I think AI is pretty smart.
03:29I don't think it's smart enough to do that. I also think by the time we get around to regulating it,
03:34or legislating on it in a meaningful way, the technology is going to be so far past us
03:39that we're... I mean, there's very few times, and I can say that up here, because the Music
03:43Modernization Act became law before I got here, and it's actually worked. It's not perfect in the
03:50new environment, but it was originally started to be written under Napster and LimeWire and all of
03:55those things, and it still exists. But I mean, what different enforcement challenges do you see
04:00moving forward in this space, Mr. Ferguson? So the commission's got its suite of statutes
04:06that it enforces, and most of them are relatively old, and they're very old compared to AI.
04:11And the stuff that you're talking about involve these massive trade-offs between
04:16promoting the innovation... I mean, look, AI will be a super important tool for new challengers to
04:21come and challenge big tech incumbency, and we don't want regulators sort of strangling that
04:26in the cradle early on. At the same time, tremendous risks are associated with AI. But
04:31these are like huge political sort of national economic trade-offs that Congress needs to make
04:36afresh. It can't be that the agencies take these statutes from the 1910s and 1930s and tries to
04:42sort of fit AI into that system. Congress has to tell us, or whichever agencies they're going to
04:48empower to do this enforcing, how to do these trade-offs. Well, and I think one of the first
04:52questions we have to answer, which is really actually hard, is how do we define open source?
04:57Open source is very different now than it was five years ago. I mean, if you can get a song
05:02online and feed it into an AI generator, I can't play that song at my business without paying
05:08royalties. But once I put it into an AI generator, I don't know. And I mean, so I understand the
05:12regulatory side of it. I just... I'm very concerned about how we actually... You all do it, and we do
05:19it, how you actually figure out a mechanism to enforce it. Because stealing something and making
05:24something not original your own is... I mean, we have the rules in place. We just don't... I don't
05:30know if we're... I think we're massively inadequately prepared to enforce it. And with
05:34that, I yield back. Okay. All right. Now...