🛡 Anthropic Specialist: Risk of Humanity's Extinction from AI Exceeds 10%
In response to the departure of researcher Jack Cox, leading alignment specialist Evan Hubinger wrote on X that he personally believes the probability of humanity's extinction from AI over ten years is over 10%, and admitted that the company currently has no plan to solve superintelligence alignment.
🌍 A rare public quantitative assessment of existential risk from a key safety figure at a leading lab strengthens the arguments of regulators and investors in favor of independent audits and funding for safety research.
👤 This is Hubinger's personal assessment, not the official position of Anthropic: he separately noted that current models pose a relatively low risk, and that the main concern is superintelligence from recursive self-improvement.
Source 1: https://x.com/EvanHub/status/2097497037956891126 Source 2: https://www.newsweek.com/anthropic-researcher-quits-warns-ai-could-kill-everyone-12418798