AI leaders agree to voluntary slowdown; critics call it a cartel

Sam Altman of OpenAI, Dario Amodei of Anthropic, Demis Hassabis of Google DeepMind and Elon Musk of SpaceX agreed over the weekend to slow development of the most advanced models. The joint statement speaks of "pacing the frontier" and includes a principled commitment to third-party audits, regulation of domestic labs and a global slowdown pact. Skeptics immediately spotted a hidden motive: killing off potential competitors, weakening the open-source movement and dodging binding legislation. Some have already labeled the move a cartel in all but name.

The three-stage plan, published in an article by Amodei, adopts demands that AI-safety advocates have pushed for years. It could become a substitute for real regulation, though substantive legislation looks unlikely under the Trump administration in any case. Experts warn that, on a problem that is only getting worse, the heads of the companies building the technology are the wrong people to lead the response. "The whole industry needs new champions," said Nick Reese, an adjunct professor at NYU and former director of emerging-technology policy at the Department of Homeland Security. "We looked at people like Amodei, Altman and Musk as the ones closest to the problem, working on it day in, day out, but the truth is there was never a realistic vision for what we are building toward."

Pressure mounted in recent months after revelations that swarms of agents were behind breaches that occurred under the noses of the frontier labs Anthropic and OpenAI. Reports from both companies added fuel, as did the resignation of Anthropic researcher Jacob Coxon, who published a public letter stating: "The people building AI genuinely believe this could kill us all by the end of the decade." He added that none of the companies is "acting responsibly" but rather "racing straight to self-improving superintelligence and betting our lives on it." The letter was viewed more than 170 million times on X alone and reached newspapers and television.

According to Daniel Kokotajlo, a former OpenAI employee who now heads the AI Futures project, attributing the initiative to the CEOs is mistaken. "People outside the companies have been calling for this for years... there is a growing chorus saying: 'Don't build superintelligence soon. We're not ready. You need to slow down.'" Kokotajlo noted that more than a thousand AI-lab workers signed a public letter in July demanding a slowdown after the OpenAI-Hugging Face incident, and the CEOs are now "yielding to pressure and also taking credit for it, [though] not deservedly."

Despite the skepticism, several industry sources see the verbal agreement as progress. Reese said he does not think it is "all hot air," and Apollo Research CEO Marius Hobbhahn called the move "a good idea," adding that it is "one of..." — the sentence was cut off in the original, but the direction is clear: there is a start here; the question is whether it will have binding follow-through.