Anthropic’s new September 2026 Threat Intelligence Report warns that AI is rapidly becoming more than just a tool for malicious actors .

The report highlights cases involving:

 Cyberattacks: AI used for reconnaissance, credential theft, malware development and adapting attacks after detection.

 Surveillance: AI helping process massive datasets, identify targets and generate detailed profiles.

 Information warfare: Fake personas, propaganda and influence campaigns being generated and scaled using AI.

 Weapons: Actors linked to China, Russia and Yemen reportedly used Claude in activities related to missiles, drones, electronic warfare and other defence applications.

 Fraud: One operation created 4,700+ AI dating personas that interacted with more than 25,000 people.

 AI model theft: Anthropic says several China-based labs attempted large-scale “distillation” campaigns to extract capabilities from Claude. One campaign allegedly generated nearly 3 million exchanges per day.