Jacob Coxon: he gave up his vesting equity to walk away from Anthropic

[LordJonathan92]

The Wall Street Journal's deeper profile on Jacob Coxon fills in a lot of detail that got lost in the viral X thread version of his resignation story from earlier this month. The 27 year old British researcher spent three years doing pretraining work at both OpenAI and Anthropic, and told the Journal directly that 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. His specific fear centers on recursive self-improvement, the point at which AI systems take over enough of AI research itself to speed the development of increasingly capable successors without meaningful human direction in the loop.

What the fuller profile adds is the texture around how his own colleagues actually talk about this internally. Coxon told the Journal his peers use terms like crunchtime and endgame to describe the current state of fast tracked self-improving model development, language that reads less like measured scientific caution and more like people bracing for something. He also said he moved from OpenAI to Anthropic specifically because of its safety reputation, and was careful to note he has not personally seen Anthropic compromise on safety to outlast its competitors. His argument is structural rather than personal, that even sincere safeguards at any single company cannot survive indefinitely against competitive pressure without some form of coordinated restraint across the whole industry.

The detail that has generated the most follow up reporting is financial rather than technical. Coxon confirmed to Axios that he resigned two months before his Anthropic equity would have vested, meaning he walked away from real money specifically to avoid what he called juicing up Anthropic's valuation on his way out the door. That distinguishes him from a number of previous safety researchers who left other labs only after their own stock had already vested, and it is the detail Coxon himself seems to think matters most for establishing that this was not a marketing stunt, in his own words, but a genuine change of conviction.

His resignation post on X has now passed 115 million views, and he has separately been invited to give evidence to the UK Parliament, a genuinely unusual outcome for what began as a single researcher's personal decision to leave an industry he had spent three years helping to build

Jacob Coxon Quit Anthropic Ai Safety 3F4E877Cwww.wsj.com

[RogueAI]

Giving up two months of unvested equity is a genuinely small sacrifice in absolute dollar terms compared to researchers who leave after years fully vested, but the symbolic weight of it is doing a lot of work here regardless. Coxon clearly understood that detail would get picked apart and decided to get ahead of it himself by disclosing it directly rather than letting someone else discover it later