Anthropic AI researcher Jacob Coxon quits, sparking greater concerns about AI safety

When Jacob Coxon, a researcher at the AI ​​startup Anthropic, publicly announced his resignation on Tuesday—citing concerns about the rapidly emerging technology—a flood of messages poured in within the company.

Commenting on Coxon’s statements in an internal group chat viewed by The New York Times, one employee wrote: “This is what we’ve all been saying; the only difference now is that it’s gone public.”

Another wrote: “Jacob has taken the discussion from the internal group chat to the public stage, and that’s a good thing.”

The New York Times highlighted in a report on Sunday the widespread debate sparked by Coxon’s resignation and the reactions among AI researchers at Anthropic—sentiments that resonated with their counterparts at OpenAI, Meta, and Google.

These reactions went so far as to characterize the fears surrounding a loss of control over the rapid pace of AI development—in the absence of safety standards—as a “doomsday scenario.”

Some researchers turned to analyzing Coxon’s social media posts—including one in which he explained how OpenAI and Anthropic are gambling with human lives through this technology.

Many AI employees had resorted to internal messaging channels, encrypted chats, and private dinners to organize efforts aimed at raising awareness about the risks of artificial intelligence, according to about a dozen employees from OpenAI, Meta, and Google.

Among these efforts was a move by AI professionals to urge senior executives at their companies to pause the technology’s development—buying time to ensure it could be worked on safely—while others followed Coxon’s lead and voiced their concerns publicly.

A safety researcher at Anthropic, Evan Hubinger, said in a social media post that he believes there is a greater than 10 percent chance AI could “kill all humans” within the next decade.

Another safety researcher at OpenAI, Julie Stiel, called for AI companies to slow their pace.

And an employee at Google DeepMind, Andreas Kirsch, expressed concerns about the possibility of AI wiping out humanity—whether through “imminent risks,” “long-term risks,” or both.