Anthropic AI Researcher Jacob Coxon Resigns, Amplifying Industry-Wide Safety Concerns

When Jacob Coxon, a researcher at the AI startup Anthropic, publicly announced his resignation on Tuesday due to concerns about rapidly emerging technology, internal communication channels at the company were immediately flooded with messages. In an internal group chat reviewed by The New York Times, one employee remarked, “This is what we’ve all been saying; the only difference now is that it’s gone public,” while another added, “Jacob has taken the discussion from the internal group chat to the public stage, and that’s a good thing.”

A report published by The New York Times on Sunday highlighted the widespread debate and reactions sparked by Coxon’s resignation among researchers at Anthropic, sentiments that found strong resonance among their peers at OpenAI, Meta, and Google. These internal reactions even characterized fears regarding a potential loss of control over accelerated AI development without adequate safety standards as a “doomsday scenario.” Some researchers closely analyzed Coxon’s social media posts, including one statement where he asserted that OpenAI and Anthropic are gambling with human lives through this technology.

According to approximately a dozen employees from OpenAI, Meta, and Google, many industry workers have relied on internal messaging channels, encrypted chats, and private dinners to organize awareness campaigns concerning the risks of artificial intelligence. These efforts have included urging senior executives to temporarily pause technology development to ensure safety protocols can be established, alongside individuals following Coxon’s example by voicing their worries publicly.

Concerns have also been shared by other prominent figures in the field. Evan Hubinger, a safety researcher at Anthropic, stated on social media that he believes there is a greater than 10 percent chance AI could “kill all humans” within the next decade. Similarly, OpenAI safety researcher Julie Stiel called for companies to slow their pace, and Andreas Kirsch, an employee at Google DeepMind, voiced apprehensions about AI potentially wiping out humanity through imminent risks, long-term risks, or both.