In May 2024, 27‑year‑old Jacob Coxon walked out of Anthropic, citing a looming AI catastrophe that, he warned, could claim humanity by the decade’s end.
Coxon, a Cambridge mathematics graduate, began his career at OpenAI in 2023, where he specialized in pre‑training large language models.
He moved to Anthropic in May, hoping the lab’s transparency would ease his fears, but incidents like the Hugging Face breach at OpenAI left him uneasy.
During a recent meeting with senior staff, Coxon voiced that even safety work felt complicit, and after consulting former OpenAI safety colleague K. Kokotajlo, he decided to quit. He told the team, “Even working on safety felt like being complicit in the race.”
His departure has rattled the AI community; Anthropic’s chief safety officer, whose name is not disclosed, responded with a statement urging continued research caution.
The United Nations has already scheduled a special session on AI safety for June 15, 2024, in response to growing fears sparked by Coxon's warning.