Two Princeton computer scientists, coauthors of "AI Snake Oil," have responded to Anthropic's claim of a >10% chance AI could kill all humans within ten years. Jacob Coxon, who quit OpenAI and Anthropic, stated both companies knew AI could cause human extinction by decade's end, a claim viewed by 156 million people. Evan Hubinger, an Anthropic AI safety lead, agreed with Coxon's assessment, saying they believe AI could kill all humans within a slightly longer timeframe and lack a plan for superintelligence alignment.

The primary risks involve destructive cyber attacks on critical infrastructure like water and power systems. Sayash Kapoor and Arvind Narayanan, known for addressing AI hype, argue alignment alone is insufficient and propose a layered safety approach including Control (sandboxing, monitoring), Downstream defence (AI defending against AI), and Resilience (recovery plans). They believe urgent action can reduce risks and shift cybersecurity balance toward defenders.