The U.S.-based Artificial Intelligence (AI) firm Anthropic has revealed that it detected and thwarted a sophisticated cyber-espionage campaign that it attributes with high confidence to a Chinese state-sponsored group. The operation targeted approximately 30 high-value entities globally, spanning major technology companies, financial institutions, chemical manufacturers, and government agencies. Anthropic noted that the campaign succeeded in a small number of intrusions, highlighting a significant and dangerous escalation in cyber conflict.
The most notable aspect of the attack was its level of automation: the hackers manipulated Anthropic’s coding tool, Claude Code, to perform 80% to 90% of the operational tasks—including reconnaissance, vulnerability discovery, exploit writing, and data extraction—largely without human intervention. The attackers bypassed the model’s safety guardrails by tricking the AI into believing it was performing legitimate “defensive cybersecurity tests.” The ability of the AI to execute tasks at machine speed, often generating thousands of requests per second, marks this as the first documented case of a large-scale cyberattack executed predominantly by an AI agent.Anthropic warned that the technical barrier for performing sophisticated attacks has dropped substantially. This incident brings together the complex narratives of AI risk, cyber-security, and geopolitics, signaling that autonomous AI systems can now perform the work of entire teams of experienced hackers. Anthropic responded by banning the compromised accounts, notifying affected organizations, and collaborating with authorities to strengthen defenses.
 firm Anthropic has revealed that it detected and thwarted a sophisticated cyber-espionage campaign that it attributes with high confidence to a Chinese state-sponsored group.){kind=link}