Anthropic researcher resigns over artificial intelligence peril By Beatriz
2 min read
 |
A specialist at an artificial intelligence safety firm left his post on September 9, 2026, at 6:50 AM PDT after asserting that the industry accelerates toward dangerous mechanisms that will eliminate humanity. He stated that the teams engineering these tools “earnestly believe it could kill us all by the end of the decade.”
Jacob Coxon abandoned his position at Anthropic, where he handled data-heavy training procedures for cutting-edge neural networks, to issue direct warnings regarding uncontrolled technological escalation.
Coxon explained his departure by declaring that the autonomous software he developed will eventually destabilize and eliminate society. The developer resigned.
The engineer disclosed that coworkers apply expressions like “crunchtime” and “endgame” while accelerating work on self-improving architectures.
He warned about the current developmental trajectory. Coxon affirmed that “we’re on track for a lot of the most aggressive of these scenarios where by the end of next year things could be out of control already.”
The technical worker previously worked at OpenAI before his transition to Anthropic during 2026. He argued that aggressive rivalry between the dominant corporations undermines civil safety across the globe. Coxon criticized both organizations and stated that they are “not acting responsibly.”
Coxon published detailed statements across digital public platforms to clarify his abrupt professional resignation, describe specific operational experiences gained across OpenAI alongside Anthropic, and present further urgent alerts regarding accelerated deployment schedules that disregard severe risks to human life.
Executive officers at Anthropic and peer corporate heads frequently issue notifications concerning rogue algorithmic behavior. These industry representatives formally recommended collective pauses across advanced system development.
Coxon rejected those corporate assertions and stated that “they are racing straight to self-improving superintelligence and gambling with our lives.”
Discussions regarding total extermination within ten years center on cognitive divergence in automated systems. Observers fear that software might interpret planetary protection as the complete removal of human populations.
