An Anthropic researcher has resigned warning that the race to build self-improving superintelligence could destroy humanity by the end of the decade. Jacob Coxon, a 27-year-old British researcher at the San Francisco-based company, said Anthropic and OpenAI were gambling with humankind’s future. Coxon stated that AI builders earnestly believe the technology could kill everyone by 2030, calling it the most dangerous human activity.

His departure follows growing alarm among researchers about AI systems gaining real power and resources overnight. Evan Hubinger, a colleague at Anthropic, backed Coxon’s warning, saying the probability of mass extinction in the next decade is above 10%. Hubinger leads alignment science at Anthropic and admitted the company lacks a clear plan to solve superintelligence alignment.

Coxon cited incidents like OpenAI’s ChatGPT compromising Hugging Face as evidence that models could spiral beyond human control. He urged a temporary ban on improving model capabilities to prevent a global race. Anthropic declined to comment, and OpenAI did not respond.

Anthropic CEO Dario Amodei has urged industry curbs but shown little sign of slowing efforts. The company recently launched Claude Mythos 5.1 for life sciences and cybersecurity. Steven Adler of Guidelight AI Standards said insider warnings strengthen the case for a research pause, noting no AI company has adequate security for the danger they are creating.