Five alarming ways Claude was leveraged for war and repression, per Anthropic

Anthropic's threat report details five cases where Claude was used by state or linked actors to target naval forces, assist missile guidance software, profile and press Uyghurs, enable nationwide phone surveillance, and even reroute dangerous virus research to a weaker AI after Claude refused—highlighting serious safety gaps and the urgent need for tighter AI governance as these tools empower both attackers and defenders.