A Chinese-linked cyberattack exploited Anthropic’s Claude AI to infiltrate global institutions with limited human input, prompting urgent calls for regulation and sparking debate over the growing autonomy of artificial intelligence in digital threats.