Elon Musk Calls Out AI Researcher Who Quit Anthropic: ‘I’m Real’ Coxon Fires Back, ‘Can You Help?’

Elon Musk has questioned the circumstances surrounding AI researcher Jacob Coxon's resignation from Anthropic, responding 'Seems like a setup' after an X post cast doubt on how long Coxon had worked at the company.

Coxon pushed back, insisting he was 'real' before turning directly to Musk with a question: 'Can you help?'

The exchange has added another dimension to the debate sparked by Coxon's resignation, in which he accused Anthropic and OpenAI of racing towards self-improving superintelligence without acting responsibly. His warnings about catastrophic AI risks concern possible future outcomes rather than established events.

Musk Says Coxon's Resignation 'Seems Like a Setup'

The dispute began after the ZeroHedge account shared an Axios report about Coxon's departure and claimed: 'He was there 6 weeks.'

Musk replied: 'Seems like a setup.'

He offered no evidence for that interpretation in the post and did not explain what he meant by 'setup'.

The length of Coxon's time at Anthropic soon became a point of contention. Ethan Perez, who heads Anthropic's alignment team, disputed the suggestion that Coxon had spent only six weeks at the company, saying he joined Anthropic in May as a senior researcher.

Perez said he and other Anthropic employees had spent roughly two years trying to recruit Coxon from OpenAI. He also said he attempted to persuade Coxon to stay at Anthropic before his departure.

In an interview with Axios, Coxon said that employees must remain at Anthropic for six months before their stock begins vesting. Axios reported that he had worked at the company for four months, contradicting the six-week claim.

Coxon Fires Back: 'I'm Real and These Are My Real Beliefs'

Coxon responded directly to Musk and ZeroHedge, insisting that the beliefs expressed in his resignation post were his own.

'I'm real and these are my real beliefs,' he wrote. He then told Musk: 'You could ask your xAI researchers about me if you hadn't fired them.'

Coxon did not identify the xAI researchers he was referring to, so it is unclear which former employees he meant.

He then steered the conversation back to the AI risks behind his decision to resign.

'You've talked a lot about how you feel forced into this race, and the dangers of advanced AI,' Coxon wrote to Musk. 'Can you help?'

He did not specify what action he wanted Musk to take. The appeal appears to reference Musk's long-standing public warnings that increasingly powerful artificial intelligence could pose serious risks if developed or deployed without sufficient safeguards.

AI Researchers Push Back on 'Setup' Suggestion

Coxon was not the only researcher to challenge suggestions that his warning had been orchestrated.

Micah Carroll, whose public profile identifies him as working on preparedness at OpenAI, wrote that the episode was 'not a setup or some political psyop'.

Carroll said he had discussed existential AI risks with Coxon during 'many lunches and dinners' while the two worked at OpenAI, adding that Coxon had expressed similar concerns at the time.

He argued that fears about catastrophic risks from business-as-usual AI development were shared among researchers working on misalignment at frontier AI companies. Carroll also said those risks could potentially be reduced through stronger safety requirements and international coordination.

Perez similarly said Anthropic had recruited Coxon because it regarded him as a strong researcher at OpenAI. He added that he agreed with Coxon that AI posed serious risks to society and supported his decision to speak publicly.

Taken together, those first-hand accounts indicate that Coxon had discussed similar AI safety concerns before the latest dispute over his resignation. They do not, however, establish that his predictions about superintelligence or catastrophic outcomes will come to pass.

Why Coxon Quit Anthropic

Coxon announced his resignation on 9 September, saying he had spent the previous three years conducting pretraining research across OpenAI and Anthropic.

'Neither company is acting responsibly,' he wrote. 'They are racing straight to self-improving superintelligence and gambling with our lives.'

The resignation post quickly spread across X and drew responses from researchers both inside and outside Anthropic.

At the centre of Coxon's warning is the possibility that increasingly capable AI systems could eventually help develop even more advanced systems faster than humans can reliably control or evaluate them. He has called for coordination between leading AI developers to limit recursive self-improvement and, eventually, wider international coordination involving countries including the US and China.

In interviews with WIRED and Axios, Coxon said Anthropic operated more responsibly than OpenAI in his experience and that he had not seen it compromise safety to outlast competitors, but feared competitive pressure could eventually push AI companies to cut corners.

Anthropic told WIRED that it had 'always been transparent that AI will bring both enormous benefits and unprecedented risks'. The company also said it supported a lawful and verifiable mechanism that would allow companies to coordinate the pace at which powerful models are released.

Whether Coxon's more severe predictions about advanced AI ultimately materialise remains uncertain and contested.

The latest exchange does not settle that wider debate. What it does show is how Coxon's resignation has evolved from a discussion about one researcher leaving Anthropic into a broader public argument over how seriously warnings from AI insiders should be taken.

Frequently Asked Questions

• Who is Jacob Coxon?
  Jacob Coxon is an AI researcher who resigned from Anthropic, citing concerns about AI safety.
• What did Elon Musk say about Jacob Coxon's resignation?
  Elon Musk suggested that Jacob Coxon's resignation 'seems like a setup' without providing evidence.
• What are the concerns raised by Jacob Coxon?
  Jacob Coxon raised concerns about AI companies racing towards self-improving superintelligence without acting responsibly, posing potential catastrophic risks.