Jacob Coxon says Anthropic conversations about 2027 shaped his exit

Interviews published on October 5, 2026, with eight former employees of OpenAI, Anthropic and Google DeepMind described different reasons for leaving—not one shared objection to AI risk. The newest specific account came from Jacob Coxon, who said informal conversations with Anthropic research leaders about 2027 helped prompt his resignation.

Their experiences raise a practical question: can researchers have more influence inside a major AI lab, or by leaving it? The interviewees reached different answers, shaped by their work and circumstances.

Eight former employees describe different reasons for leaving

The accounts span concerns about AI development, a desire for more say in company decisions, disagreement over defense work, the effects of AI on jobs, organizational changes and research fit. Not every departure was about existential risk. Alex Turner, for example, said that was not why he left Google DeepMind.

That range matters. A researcher can leave over a specific contract decision, feel their influence inside a company has faded, or decide that work on AI’s effects on people belongs outside a lab. Those are related to the broader debate over AI, but they are not interchangeable motives.

What Jacob Coxon said about Anthropic’s 2027 plans

Coxon worked at OpenAI from 2023 to 2026 and at Anthropic from May to September 2026. He said conversations with Anthropic research leaders had treated 2027 as a period when events could become unusually consequential. The people he spoke with, he said, did not expect government regulation or international cooperation to arrive in time.

Coxon described informal discussions, not an official Anthropic forecast. He said those conversations helped prompt his resignation. He also speculated that a future model generation could become as capable as people at coming up with research ideas. That was a projection about possible future capabilities, not a demonstrated result.

When leaving seemed more influential than staying

Daniel Kokotajlo said he left OpenAI after losing confidence in its leadership and wanting more freedom to speak publicly and publish research. For him, working outside a lab offered a better route to public advocacy.

Alex Turner described a different conflict. While at Google DeepMind, he said, he organized internally and proposed alternative language for the company’s response to a Department of Defense contract. He said the contract was ultimately signed and that the dispute informed his decision to leave. Google DeepMind responded that it had reviewed his framework, listens to employees at every seniority level, and that Turner did not understand the company’s work in that area.

Turner’s account, unlike Kokotajlo’s, centered on a specific disagreement over defense work. His departure was not a general statement that everyone should leave a lab—or that existential risk was his reason.

Audits, labor concerns and changing priorities

Miles Brundage said he left OpenAI in 2024 after concluding that his influence inside the company was diminishing and that his executive role limited what he could say publicly. He later founded AVERI, a nonprofit he described as conducting third-party audits of AI companies.

Brundage described audits that could combine model testing with reviews of internal processes and evaluations, documents and staff interviews. The approach would examine more than a model’s outward behavior: it could also look at how a company assesses and manages its systems.

Pamela Mishkin said she found it harder to justify doing work on AI’s effects on people and jobs inside a lab than outside one. She had spent a long time at OpenAI, and her account centered on where that research could be more useful.

Jeff Wu described multiple factors behind his OpenAI departure rather than a single trigger, including the company’s direction and the fit of his research. Rosie Campbell said her policy-frontiers team was dissolved after her manager, Brundage, announced his departure at the end of 2024. She left after she could not find another team doing the work she wanted.

Why some researchers stay inside AI labs

Leaving is not the only way to try to influence AI development. Mishkin pointed to the money, computing resources and information available inside labs, as well as the possibility that researchers can make a difference there. Jacob Hilton also described lab jobs as a strong career path, offering pay, prestige and experience with frontier research.

Those advantages can make a move outside a lab difficult, particularly for early-career researchers. Hilton’s account highlights the trade-off: outside work may offer more room to speak or pursue a different agenda, while lab work can provide resources and career opportunities that are harder to find elsewhere.