Adam Khoja He is a researcher at the Center for AI Safety, where he works on AI safety benchmarking, forecasting, and the geopolitics of AI. He was part of the team that b...

“we're probably at the point of capabilities where it would be in our interest to be slowing the rate of capabilities progress.”

Doom Debates! · AI Research & Frontier Labs · September 2026

Context

Liron Shapira asks Khoja directly whether he personally supports pausing AI. Khoja, who works on AI safety at CAIS, separates the two: slowing is already warranted, a full pause he is not ready to call for yet. He adds that the point where a slowdown becomes clearly necessary may arrive quickly.

Listen to the episode → More from Adam Khoja → More AI Research & Frontier Labs quotes →

Transcript

Doom Debates! Around 1:14:42 into the episode

Adam Khoja

…this political will could appear eventually. But for the sake of the article, we're just assuming that that has appeared one way or the other. And so a lot of sort of more drastic actions that you otherwise wouldn't consider become possible or on the table if you find yourself in that scenario.

Liron Shapira

Would you personally want to see us pull the trigger on some kind of slowdown agreement very soon? Or another way to ask it is: do you support pause AI?

Adam Khoja

Yeah, so I think that in general, we're probably at the point of capabilities where it would be in our interest to be slowing the rate of capabilities progress. I don't know if fully pausing right now would be like the best approach, though I think it's like obviously a really complicated topic and it's going to be contingent on a bunch of sort of fiddly empirical upstream variables that where sort of our assessment is changing every day. But I do definitely think that we may quickly reach the point where a slowdown becomes clearly necessary.

Liron Shapira

Mutual spying sounds great to me. I feel like that even would be something we could have backported to the nuclear arms race, right? Because from what I've read, the US really overestimated, or at least they conveniently or rhetorically overestimated what Russia was doing. Like the U.S. was…

Adam Khoja

Yeah, I mean, I have news for you, which is that we're all like already spying on each other. And so the point that we make in the article is like this may not actually be that much different in terms of IP leakage than the status quo. So, I mean, you could talk about, for instance, China, for…

Liron Shapira

Yeah, I mean, I think this is a great article because so many people that I talk to, their first reaction is like, you're being so unrealistic. China's never going to cooperate. And it's like, really? We just have to agree to spy on each other and call each other out. And nothing wrong with also…

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

More from Doom Debates!

“I think the number one takeaway here is that this is the final nail in the coffin of the Ed Zitron camp, right? This idea that, like, look, sure, Anthropic is making tokens, but they're just borrowing from NVIDIA, right? And all these companies are just like in a circle, like throwing money at each other, and there's no value being generated. There were so many data points from the Anthropic S1 that put the nail in the coffin of that kind of speculation. So, number one is just like the cash flow. It was very clear that when Anthropic is serving a model, the margins on serving the model are huge. They were talking like 50, 60, 70, 80%, like depending on how you account for it. But like, yeah, they serve a model. They're printing money. Like, serving the model is printing money.” — Liron Shapira, Doom Debates!

“As of a year ago, I had this theory when people were saying, where's the productivity growth? … But I had this theory that what was actually happening was that the workers who had figured out how to use AI weren't telling their bosses. They were just spending their day at the beach, right? And so they're still getting as much money as they were to do the same amount of work they were doing. So the productivity growth doesn't show up unless you count beach time.” — Robert Wright, Doom Debates!