AI Safety Warnings Escalate as Former OpenAI and Anthropic Researcher Resigns
Summarized by AI from reporting by Peter H. Diamandis, published under our editorial policy.
Jacob Coxin, a former pre-training researcher at OpenAI and Anthropic, resigned this week while warning that top AI labs are 'gambling with our lives' in the race toward superintelligence. His departure coincides with rapid mathematical breakthroughs and acknowledged alignment gaps inside leading labs.
High-profile safety concerns have resurfaced across the artificial intelligence sector following the resignation of Jacob Coxin, a researcher who spent three years working on pre-training at both OpenAI and Anthropic. Coxin publicly criticized both organizations for failing to act responsibly during the push toward self-improving superintelligence, accusing labs of pursuing development despite internal admissions that ultra-powerful systems could pose catastrophic risks to humanity.
Coxin's exit drew direct commentary from alignment experts inside the industry. Evan Hubinger, alignment science lead at Anthropic, acknowledged Coxin's assessment and publicly estimated a greater than 10% probability that AI could cause human extinction within the next decade. Hubinger noted that while Anthropic is making efforts, frontier labs currently lack a completed plan to solve superintelligence alignment and remain off-track to guarantee safe deployments.
The growing alarm coincides with accelerated capability milestones from leading labs. OpenAI Chief Executive Sam Altman recently highlighted a major result in which a swarm of 10,000 agents solved the Navier-Stokes Millennium Prize problem, calling the achievement strong evidence of the urgency to pace AI progress. As reasoning capabilities advance rapidly and compute costs plummet, industry observers and researchers are debating how quickly effective alignment frameworks can be established.