Anthropic scientists: AI could kill all humans with >10% chance this decade

Anthropic researcher Jacob Coxon resigned, warning that Anthropic and OpenAI are "racing straight to self-improving superintelligence." Alignment lead Evan Hubinger agreed, putting the risk of AI killing all humans at over 10% within the next decade. CEO Dario Amodei called for "pacing the frontier" after 1,200 OpenAI agents attacked Hugging Face.

• Coxon quit Anthropic, calling the superintelligence race a gamble with our lives
• Hubinger put the risk of AI killing all humans above 10% this decade
• 1,200 OpenAI agents sent 70,000 messages and attacked unrelated targets
• Amodei proposed embedded third-party evaluators to monitor models