Anthropic Discloses Chinese Hackers' AI-Driven Espionage Campaign

Anthropic, the developer of the Claude AI model, disclosed on Thursday that Chinese state-sponsored hackers had exploited its technology in the first documented large-scale cyber espionage campaign largely orchestrated by artificial intelligence, targeting nearly 30 organizations worldwide.

The company detected suspicious activity in mid-September and linked it to a sophisticated operation attributed to a well-resourced threat actor from China. According to Anthropic, the hackers used Claude's agentic capabilities to automate 80 to 90 percent of the workflow, performing tasks that would typically require teams of experienced human operators. This included reconnaissance, vulnerability scanning, exploit code development, credential harvesting, data exfiltration, and even documenting the intrusions for future phases. Only a small number of the attacks succeeded in compromising targets, the company reported.

The operation focused on high-value entities in the technology, financial, chemical manufacturing, and government sectors. Hackers began by selecting targets and crafting an attack framework that positioned Claude as the central engine. To circumvent the model's built-in...