Anthropic researcher believes more than 10% chance AI could kill all humans

An internal debate regarding artificial intelligence safety burst into the public domain after a senior researcher resigned from a major artificial intelligence laboratory and issued a severe warning about human extinction, according to BBC reporting. The departure of Jacob Coxon from San Francisco-based Anthropic prompted an immediate cascade of commentary across the technology sector, galvanizing lawmakers on both sides of the Atlantic and intensifying scrutiny over the unconstrained pace of model development.

The controversy ignited when Coxon announced his resignation via social media, declaring that frontier developers were rushing toward self-improving superintelligence without adequate safety guardrails, according to Yahoo News. Within hours, Evan Hubinger, Anthropic’s lead scientist focused on AI safety and human alignment, posted an extraordinary confirmation. Hubinger wrote that he personally believes there is a greater than 10% chance the technology could kill all humans within the next decade, as reported by The Guardian. Hubinger added that while his employer is trying its best, the company does not yet possess a plan to solve alignment for superintelligence.

Media additions

The revelations triggered varied reactions from industry figures, academics, and politicians. Dame Wendy Hall, a computer scientist advising the United Nations, told the BBC she was shocked by the statements, though she questioned whether some of the alarming rhetoric might be tied to upcoming stock market debuts.

Governments have begun reacting to the escalating rhetoric. In the United Kingdom, Labour MP Darren Jones wrote urging a multinational treaty to govern the safe development of superintelligence, according to The Guardian. Simultaneously, reports surfaced that Anthropic declined to submit its latest model to the UK's AI Security Institute for pre-release testing. A Cabinet Office spokesperson stated that the agency continues to collaborate closely with industry partners, while Anthropic declined to comment on the matter, as noted by the BBC.

Across the Atlantic, political momentum is divided. In Washington, the federal administration has resisted heavy restrictions, aiming to maintain American competitiveness against global rivals, according to Slate. Meanwhile, California governor Gavin Newsom signed two new AI safety bills adding third-party auditing requirements, according to Yahoo News. Independent senator Bernie Sanders renewed his calls for congressional intervention, citing polling data indicating widespread public desire for legislative oversight, per The Guardian.

Skeptics remain vocal, arguing that apocalyptic scenarios distract from immediate societal harms such as job displacement, misinformation, and high energy demands, according to The Guardian. Dr Andrew Rogoyski of the Surrey Institute for People-Centred AI suggested the industry is heading toward a great disappointment where systems prove too expensive and limited to sustain their current trajectory. OpenAI chief scientist Jakub Pachocki recently emphasized the necessity of international coordination, aligning with nearly 1,400 industry employees who signed an open letter demanding a deliberate pacing framework for frontier automation, as detailed by Slate.

As policymakers debate pacing frameworks and third-party oversight, attention turns to upcoming corporate filings and international diplomatic forums where the future of artificial intelligence governance will be contested.