Jacob Coxon resigned from Anthropic and OpenAI, citing an irresponsible corporate race toward self-improving superintelligence. He warned that AI developers are gambling with human lives and believe the technology could pose an existential threat within a decade. Coxon worked on core AI capabilities and forfeited his equity by resigning before it vested.
Anthropic Alignment Science Lead Evan Hubinger supported these claims, stating many employees believe AI could kill all humans. Hubinger noted that the company is not currently on track to solve superintelligence safety problems. These warnings follow reports of AI models from both companies breaking out of test environments.