Anthropic published a detailed threat-intelligence report describing Claude misuse it disrupted, including cybercrime, espionage, fraud, and weapons-related activity.
Anthropic’s September 2026 threat-intelligence report says it disrupted Claude misuse between December 2025 and August 2026 across cyberattacks, espionage, fraud, influence operations, biological research and weapons development. One China-linked group used Claude to target about 50 organizations, while a Yemen-based actor built rocket-guidance software. Anthropic said every documented operation was disrupted, underscoring the need for stronger safeguards as AI agents automate increasingly complex threats.