Why Are AI Companies Talking All About Pacing? - DTNS 5352

Jason Howell (0:09)

This is the Daily Tech News for Monday, September 14th, 2026 We tell you what you need to know, give you the important context on all of it, and help each other understand.

Robb Dunewood (0:18)

Today, pacing is the word of the day in AI, with OpenAI, Google Anthropica, and Microsoft. And Justin Robert Young tells us about how AI has become the top political issue heading into the US midterms.

Jason Howell (0:29)

Yeah, let me just tell you, there's a lot of AI, but we've got other things to talk about today. I'm one of your hosts, Jason Howell.

Robb Dunewood (0:34)

And I'm Robb Dunewood.

Jason Howell (0:36)

We're going to start with what you need to know with a big story.

Yes, as you said, the word of the day is pacing. This is one of those things you're going to start hearing a lot more because, boy, did it come out in force over the weekend. Anthropics CEO, Dario Amadei, published an essay titled, We Must Pace the Frontier. And in that essay, detailed a three-part commitment and kind of a proposal to get people thinking, and boy, did it have an impact.

First of all, giving outside evaluators, these are groups like Meter, the same ongoing access that their own internal safety teams get so that they can watch the work as it happens instead of reviewing it all after the fact. Secondly, Amadei wants leading democratic AI labs to agree on shared safety standards and pacing limits, with the government refereeing those talks, if you want to call it that, rather than just joining them. And then thirdly, he wants coordination with governments, or their authoritarian governments, including China, like reaching agreement on bans for things like bioweapons research. There's a lot more in there, but pacing, this word you're going to be hearing a lot, essentially means deliberately slowing the rate at which these frontier AI models are developed, that rate at which they're becoming more capable. With that slowdown, it leaves room for more safety testing, more monitoring, controls around security, that sort of stuff. It does not mean a total freeze on AI research on releases, but it's a slowdown. It's kind of like you're setting a slower pace, is what it's all about. In response, you got OpenAI CEO Sam Altman agreeing, saying that labs should accept slower capability gains, should accept independent safety checks, so safeguards and human control are not lost along the way.

Altman also, by the way, said that OpenAI has ruled out a 2026 IPO. This is something we've been expecting to possibly happen, saying that it would be, in his words, ill-advised to do so right now.

With the timing of that, we can talk about that to this. It's interesting timing. Microsoft AI CEO Mustafa Soliman published a 37-page draft titled Humanist AI Code of Conduct. There describes requiring MAI models to remain subordinate to human control, to never resist shutdown. And then the US president pushed back on all these pacing calls, saying, quote, We can put guardrails, we can do this and that, but I think you have a lot of negative forces that are bringing it up that shouldn't be bringing it up, and they're bringing up things that won't happen. And then I know this is a lot, but we're going to talk about it in a second. This is a very evolving and complicated story, to be honest. But Financial Times reports that Anthropic is also expecting its second straight profitable quarter. Company has picked NASDAQ for a potential IPO in October. They're targeting valuation around $2.3 trillion. NVIDIA investing as much as $10 billion, and that is an anchor investor. Interesting timing here, especially paired up with OpenAI. According to the information also, Anthropic, OpenAI and Google reportedly held private talks ahead of Amadeus' weekend essay on an industry AI safety standards body that would be focused on testing and auditing these frontier models. So that was a lot. And I apologize about the wall of information there. I mean, we could go a million different directions with this. But Robb, I'm curious to know, like, what are your thoughts on this pacing discussion that is moving quickly?

Robb Dunewood (4:29)

So maybe it came to a head last week over the last couple of weeks, but there has been a fairly steady stream of developers and mathematicians and folks pretty high up in various AI companies that are leaving these companies at great financial peril. They are leaving a lot of money on the table. And they're saying things like, we could destroy Earth with this if we don't slow this down.

We could very easily create something that's going to be, it's going to recursively improve itself. Literally, you've got AI improving on AI, and it does something that ultimately could wipe out all life on Earth. And some people think, well, that's, you know, somebody playing chicken little, but there's enough of these folks who are saying these things. And not only that, the developers and the researchers and the mathematicians that are still at the companies are saying, oh yeah, they're right. This is stuff we talk about all the time.