SAN FRANCISCO, United States: Anthropic AI safety concerns intensified after researcher Jacob Coxon resigned and accused Anthropic and OpenAI of moving too quickly toward more powerful artificial intelligence.
Coxon, who previously worked at both companies, wrote on X that he left because he believed they were “gambling with our lives.” His post drew more than 70 million views, CNBC reported.
Anthropic alignment lead Evan Hubinger later backed Coxon’s broader concern. Hubinger said he personally believes there is more than a 10% chance that AI could cause human extinction within the next decade.
He also said Anthropic does not yet have a clear plan for aligning a future superintelligent system. The figure represents Hubinger’s personal risk estimate, not a measured probability or industry consensus.
Read: Bill Gates AI Limits Call Targets Jobs, Safety Risks
OpenAI chief scientist Jakub Pachocki separately wrote that AI companies have not solved alignment and monitoring well enough to continue scaling indefinitely at maximum speed.
Pachocki called for voluntary slowdowns when necessary and greater international coordination around future AI development.
The warnings come amid debate over recursive self-improvement, a hypothetical capability in which an AI system could help design more capable successors. The supplied source notes that such self-improvement is not currently possible.
Read: Astra AI Monitorability Debate Grows Over Hidden Reasoning
U.S. lawmakers have also proposed measures including the FRONTIER Act and the Ban Artificial Superintelligence Act as Congress considers how to regulate increasingly capable AI systems.
