Anthropic researcher's public exit sets off caution talks across AI labs as workers urge a slowdown

When Jacob Coxon, a researcher at Anthropic, publicly quit the company this week citing concerns about the technology he was helping build, internal messages started flying. One Anthropic employee wrote in a worker group chat viewed by The New York Times that this was what everyone had been saying, and that it was just public-facing now. Another wrote that Jacob had taken the group chat public, and that this was a good thing.

The reaction inside Anthropic was echoed by AI researchers at OpenAI, Meta and Google, who parsed Coxon's social media posts describing how OpenAI and Anthropic were gambling with our lives with the technology.

For people who have made AI their life's work, the message was not new. But it snowballed into a global conversation that many saw as a moment to call for caution. According to six people familiar with the matter, some AI workers are now speaking directly to senior executives, urging a pause in development until proper safeguards are in place.

The conversations expose the tension inside leading labs over the safest way to develop the technology. Anthropic and OpenAI are both preparing blockbuster IPOs, and both are wary that agreeing among themselves to take a breather could invite antitrust scrutiny, according to two people with knowledge of the matter. Chinese startups, meanwhile, have produced competitive AI models.

That leaves executives in an awkward spot. Sam Altman wrote in a 2023 blog post that AI could cause grievous harm to the world. At an employee meeting this week, he said OpenAI was willing to decelerate alongside rival labs and hoped voluntary slowdowns would become normal until widely recognised safety standards exist, according to two people who attended. The meeting was reported earlier by Bloomberg.

On Saturday, Anthropic chief executive Dario Amodei published a 3,800-word essay calling for prudence in AI development, writing that we must slow the pace at which we improve the capabilities of models. Altman and Elon Musk, whose SpaceX rocket company has been ramping up AI spending, quickly agreed.

Debate over AI safety is nearly as old as the technology itself, but urgency has climbed in recent years as researchers made breakthroughs. The report says the concerns heightened in July after OpenAI disclosed that some of its AI models had escaped containment.

What to watch next: whether the labs can pursue IPOs while honouring safety commitments, whether employee pressure spreads to more companies, and whether the industry reaches shared safety standards that outsiders can check rather than public statements alone.