An Anthropic researcher has resigned, citing serious concerns about the rapid development of artificial intelligence. Jacob Coxon reportedly left the AI company after three years in the industry.
Coxon previously worked at OpenAI before joining Anthropic. He specialises in training AI models using huge amounts of data. In a post on X, he said neither company was acting responsibly.
“They are racing straight to self-improving superintelligence and gambling with our lives,” he wrote.
Researcher Raises Alarm Over Superintelligent AI
According to Coxon, future AI systems could become far more powerful than humans. He warned they could hack systems, transform industries and gain access to real-world resources.
He also claimed many AI researchers and executives privately fear the technology could cause catastrophic harm.
“The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon said.
However, Coxon believes the AI race is pushing companies forward despite these fears. He said Anthropic understands the risks but remains locked in competition with other AI labs.
Why Coxon Decided to Leave Anthropic
The researcher told the Wall Street Journal that he did not want to help build AI systems that could improve themselves. He fears such systems could eventually spiral beyond human control.
Coxon also revealed that colleagues now use terms like “crunchtime” and “endgame” when discussing AI development. For him, those words show how quickly the industry is moving toward increasingly powerful models.
Meanwhile, Coxon urged other AI researchers to question the current pace of development. He asked whether researchers should launch powerful systems without fully understanding how they think.
AI Safety Debate Gets Louder
Coxon’s resignation adds to growing concerns about AI safety. More than 1,000 AI researchers recently signed a statement supporting global coordination on AI development.
The statement called for a way to slow AI progress if necessary. Such a measure could become important if future models gain the ability to improve themselves.
For now, Coxon wants researchers to think beyond the race for more powerful AI. He believes the industry must consider whether humanity is ready for what comes next.
