David Sacks, co-chair of the President's Council of Advisors on Science & Technology, says Anthropic and OpenAI should slow development of their most advanced AI models if they believe the technology poses serious risks, but argues they don’t need new regulation to do it. He pushes back against calls for antitrust exemptions and a broader regulatory framework, saying today’s frontier AI labs are setting the frontier themselves and do not need "anyone else's permission." He also underlines that any approach must account for competition with China, which is unlikely to join a global agreement to...
What David Sacks said
Written from the verified transcript and checked against it. Every figure links to the moment it was said.
David Sacks, founder of Craft Ventures, argued that Anthropic and OpenAI, which he called a duopoly at the frontier of AI, should voluntarily slow development to ensure product safety without demanding government concessions. He cited Lina Khan's statement on product liability as a key incentive, and said existing laws already cover AI harms. He rejected the need for antitrust waivers, calling such requests 'bartering.' Sacks dismissed claims that safety requires government intervention, and criticized Bernie Sanders' bill to pause AI as a threat to U.S. competitiveness with China. He advocated for transparency and audits, but questioned the independence of evaluators like those funded by Anthropic's investors. He acknowledged Dario Amodei's concern about falling behind China, but called it a contradiction, urging lab leaders to focus on engineering and reliability. He framed the safety debate as a political 'hoax' ahead of midterms.
Key takeaways
Sacks said Anthropic and OpenAI form a duopoly with a two-generation lead in model capabilities, and should slow development on their own.
He cited Lina Khan's statement that AI companies face product liability as a key incentive for safety.
He called Bernie Sanders' bill to pause AI with 20-year prison sentences a threat to U.S. competitiveness with China.
Numbers and commitments
Figure
What it refers to
Type
At
20 year prison sentences
proposed penalty for AI developers under Bernie Sanders' bill
0:00Why don't we need regulation in what those frontier labs proposed today?
3:17And in that regulatory capture concern, that's the other question is, well, should the government work in getting those people into the room to work together, or should they work to prevent that?
5:04How much does the president want the opposite for the frontier labs to keep going?
6:44Just to your mind, David, where is the most appropriate area to regulate for in the in what is now the broad field of AI?
The frontier labs say we need to pace advancement at the frontier. And in your post you said, okay, do it yourselves. But it kind of makes this central question of why don't we need a government intervention? Why don't we need regulation in what those frontier labs proposed today?
Well, the first point is, if you look at the structure of the market for frontier AI right now, we do have a duopoly situation. By any metric, whether it is market share or revenue growth or model capabilities, there are basically two companies at the frontier. And that is Anthropic and OpenAI. And you have Dario and Sam have not both said, okay, let's pace the frontier to make sure our products are safe. My point is, guys, go ahead. I think if you feel that the right priority right now in your product roadmap is to make your products more reliable, more trustworthy, more predictable, then you should do that. And by the way, they have every incentive to do this. Lina Khan just came out with a statement that I very much agree with, where she said that the major AI companies are subject to product liability for their products. So when or if they release an unsafe product, they're mostly liable for that. So this is very much in their business interest. I don't really believe that they're doing this or need to do this for altruistic reasons. I think every company in America needs its products when they release them to be safe. That's just a basic corporate responsibility. And if you don't, you're going to be subject to all sorts of civil and maybe even criminal liability, as Lina Khan was pointing out. So, you know, again, we're not bereft of laws and regulations in this country. We have many laws, statutes, regulations that apply to AI. For example, you cannot use AI to conduct a cyber attack. You can't commit fraud. You cannot commit privacy violations. There's child safety regulations. And I'd say most importantly, we have product liability rules. So again, we have plenty of regulations here. If a little frontier labs release products that are unsafe. And again I would turn it back to them. It's on you guys to make sure your products are safe. If you feel like you need to quote unquote pace the frontier to do that, go for it. Do that right. You don't need anyone else's permission. And let's make one other point on this. You know, they seem to be not there. I think Dario basically said in his blog post that he needs an antitrust waiver. Well, no, you don't, and you really don't. You should go ahead and make your products safe without demanding any concession from the government. You know, we don't believe that you should be free from competition. I'm sure every company in America would love to have that kind of deal. You have a basic obligation to make your products safe. Do not demand antitrust waivers or liability waivers in exchange for that. Honestly, it feels like they're bartering with us. And I don't think they should do that. I think they should just go ahead and do what they need to do to make sure their products are safe. And frankly, if they can't do that, if they can't make their own products safe, then I think the public is going to wonder, why are they doing it? And the public will suggest that they should step aside and make room for people who will do it safely.
David, in part through your post, but many have raised the question of regulatory capture. Right. And one thing that indirectly, Sam Altman and I am a day agree on is that this pacing at the frontier only works if all of the frontier labs agree to do it together. And in that regulatory capture concern, that's the other question is, well, should the government work in getting those people into the room to work together, or should they work to prevent that? You know, in order to you spoke about the cartel issue, for example, in your post.
Well, again, if you look at the market share, you know, for frontier intelligence, if you look at the revenue growth, you know, there's really two companies that dominated. If you look at the model capabilities, Anthropic and OpenAI are probably about to generate, what they have in the lab is about two generations for everyone else. So the others are trying to catch up. But the reality is they have a pretty substantial lead. And so if Dario and Sam both say that they're willing to slow down their development. And by the way, I don't think they're going to stop development. I think this is just a shift in priorities where they're saying they're going to focus on, let's say, reliability and predictability instead of raw power. Guys, go for it. You have every incentive to do that. You have liability to do that. And you have a pretty big lead in the market. So I don't really believe this claim that they can't make their product safe unless the government steps in and, you know, imposes requirements on companies that aren't even at the frontier. Again, this feels like an excuse to me. It feels like bartering. It feels like they are trying to demand concessions from the government in order to meet their basic corporate responsibility, and I just don't buy it.
How much does the president want the opposite for the frontier labs to keep going? I'm trying to pass what he said over the weekend with some of the posts that he's made today.
5 more exchanges in this transcript
Sign in free to read the rest of this interview. No card required.
Sacks, D. (2026, September 14). David Sacks: Anthropic, OpenAI Should Slow AI If Deemed Necessary [Interview transcript]. Bloomberg Tech. CEOInterviews.AI. https://ceointerviews.ai/interview/1348124/
MLA
David Sacks. "David Sacks: Anthropic, OpenAI Should Slow AI If Deemed Necessary." Bloomberg Tech, 14 Sep. 2026. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/1348124/.
BibTeX
@misc{sacks2026_1348124,
author = {David Sacks},
title = {David Sacks: Anthropic, OpenAI Should Slow AI If Deemed Necessary},
howpublished = {Interview transcript, Bloomberg Tech. CEOInterviews.AI},
year = {2026},
month = {sep},
url = {https://ceointerviews.ai/interview/1348124/},
note = {Speaker-attributed transcript with timestamps}
}
