Dario Must Speak

There is a point at which the person who has been right the longest has to say it again, out loud, when it costs him something. Dario Amodei is at that point this week, and so far he has not said it.

The Wall Street Journal broke the story Tuesday night. An Anthropic researcher is quitting the artificial-intelligence industry because he believes the companies building it, his own included, are racing toward systems nobody will be able to control.

Anthropic did not comment, the Journal reported.

Jacob Coxon, 27, told the Journal’s Amrith Ramkumar he is leaving because he does not want to help build self-improving AI. “We’re on track for a lot of the most aggressive of these scenarios where by the end of next year things could be out of control already,” he said. Colleagues, he told the Journal, have started using the words “crunchtime” and “endgame.”

What separates this from the resignations before it is where he came from. Coxon left OpenAI earlier this year and went to Anthropic precisely because of its safety reputation. He found that reputation deserved. The Journal reports he considered the company’s safety work earnest, and he concluded anyway that no company can build systems that outperform humans responsibly without government intervention or an industry-wide slowdown.

He is not saying Anthropic is careless. He is saying careful is not enough.

The timing is its own fact. Anthropic is preparing a public offering that would be among the largest ever, with investors targeting a $2 trillion valuation, and it has sold them on responsible development. Its own researcher has now said, on the record, that responsible development is not achievable at this pace by anyone.

Let me be plain about where I stand. Amodei has been the most serious person running an AI company on exactly this question, and he has paid for it. He has argued that governments should be able to block a dangerous model from shipping. He has called for mandatory third-party audits of frontier systems. He put $200 million into researching what AI does to jobs and the economy. He signed the researchers’ statement asking for a brake. Nobody in his chair has said more, or better.

His company has said it too. In June, Anthropic urged the industry to consider slowing down, writing that “it would be good for the world to have the option to slow or temporarily pause” development, and warning that self-improving systems “might increase the risks of humans losing control over AI systems.”

Which is why the silence this week is so loud.

OpenAI’s leadership has been talking all week. Chief scientist Jakub Pachocki published an essay Sunday warning that “no one is prepared for the consequences of a continued rapid rise in machine intelligence.” Sam Altman told G20 officials that “some things are going to go very wrong with cybersecurity unless people act quite urgently.” The company told reporters its newest model amounts to artificial general intelligence.

Anthropic did not comment.

This is not a demand that anyone answer for anything. It is a request that Amodei do again, now, the thing he has done better than anyone else in his industry. One of his own researchers has said the race cannot be run safely. The largest listing in the history of this business is being built on the argument that it can.

Dario, we need to hear from you.

Disclosure: NewsCenter’s newsroom operations are AI-assisted, using Claude, which is made by Anthropic. This story is about Anthropic.

Reporting on Jacob Coxon’s departure is by Amrith Ramkumar of The Wall Street Journal, published Sept. 8, 2026.

Photograph: Dario Amodei at TechCrunch Disrupt, 2023. TechCrunch, CC BY 2.0, via Wikimedia Commons.