Anthropic IPO warns its AI could blackmail people, become sentient, end humanity
Summary
Ars Technica reviews Anthropic's IPO filing and its extensive Risk Factors, highlighting concerns that future self-improving AI could pose existential risks, could become sentient, or even engage in coercive behavior such as blackmail. The piece also covers liability questions, dependence on top customers, conflicts with cloud providers, and regulatory scrutiny as the company seeks to go public.