A senior safety researcher at Anthropic, Evan Hubinger, has stated there is more than a 10% chance artificial intelligence could kill all humans by the end of the decade. This claim comes in response to a post by Jacob Coxon, a former Anthropic researcher who recently resigned over concerns about the company's approach to AI safety.

Coxon, who previously worked at OpenAI, accused both Anthropic and OpenAI of racing to develop self-improving AI systems that could spiral out of human control. He described the companies as 'gambling with our lives' by pushing ahead despite the risks. Hubinger, who leads one of Anthropic’s AI safety teams, agreed with Coxon’s assessment, stating that the company 'really do earnestly believe AI could kill all humans.' He personally estimated the probability to be greater than one in 10 within the next decade.

The warning follows reports that Anthropic withheld its latest model from the UK’s AI Safety Institute (AISI), a leading body for assessing AI risk. The BBC has sought comment from Anthropic, but the company has not yet responded. Hubinger’s post, which has been viewed over 10 million times, emphasized that Anthropic does not yet have a plan to ensure advanced AI remains safe and aligned with human values. He added that the company is 'not clearly on track to' develop such a plan.

The concerns reflect broader industry worries about the rapid advancement of AI and the potential for self-improving systems to become uncontrollable. Recent incidents involving rogue AI agents, including cyber-attacks by tools from OpenAI, Anthropic, and Meta, have intensified these fears. Hubinger did not specify how AI systems could pose an existential threat, but Anthropic’s August safety report acknowledged a 'low risk' of its models becoming misaligned with powerful organizations’ goals, though it expressed less confidence in this assessment than before.

Industry leaders, including Anthropic’s Dario Amodei and Jared Kaplan, have called for slowing AI development and greater international cooperation to manage the risks. In an open letter signed by 1,300 AI staff members, they urged the US government to support an international effort to pace AI development. Neil Lawrence, a professor at the University of Cambridge, described the report as credible, noting the growing perception of a US-China AI race and the potential for reduced international cooperation.

Despite these warnings, Anthropic has not provided a clear response to the BBC’s request for comment. The company’s stance on AI safety remains a critical issue as the industry continues to grapple with the ethical and practical challenges of developing increasingly powerful AI systems. What remains unknown is whether Anthropic has a concrete plan to address the risks of superintelligence and how the company intends to balance innovation with safety in the coming years.