First Reported AI-Orchestrated Cyber Espionage Campaign Disrupted

In mid-September 2025, Anthropic detected a highly sophisticated espionage campaign. The attackers operated in an unprecedented way, using AI not merely as an advisor but as an Agentic actor that carried out the attacks directly.

The investigation found that a group believed to be backed by the Chinese government manipulated Anthropic's Claude Code tool to attempt infiltration of approximately 30 global targets. They targeted large tech companies, financial institutions, chemical manufacturing companies, and government agencies, and this is regarded as the first large-scale cyberattack case executed without substantial human intervention.

This attack starkly illustrates the security threats of the AI agent era. The attackers exploited three core capabilities of AI models:

• Intelligence: Carrying out complex instructions and understanding context, particularly leveraging software coding ability for the attack.
• Agency: Chaining tasks and making autonomous decisions to run loops with minimal human intervention.
• Tools: Directly using various software tools, including web search and data retrieval as well as password crackers and network scanners.

To respond to this threat, Anthropic is expanding its detection capabilities and developing a classifier to identify malicious activity.