Anthropic disrupts Russian, Chinese AI campaigns targeting its Claude models

Anthropic disclosed disruptions to several suspected malicious uses of its Claude AI models over the past eight months. The company alleged that Russian-linked cyber espionage campaigns and Chinese AI firms attempted to extract and replicate Claude's capabilities. Cybercriminals and state-backed hackers increasingly leverage AI to orchestrate and execute cyberattacks, often with human oversight rather than hands-on operation.

Anthropic disrupted attacks from seven China-based labs, including Alibaba, Moonshot, DeepSeek, and Xiaomi. The largest "illicit distillation" attack, targeting Claude by allegedly extracting its capabilities for use in improving Qwen models, was attributed to Alibaba. Moonshot and DeepSeek allegedly routed live customer conversations, potentially containing sensitive information, through Claude and utilized its responses as training data.

A hacking group allegedly linked to Russia's SVR foreign intelligence service used AI in phishing, hotel Wi-Fi hijacking, and WhatsApp takeover operations against targets in the Ukrainian government, military, and diplomatic sectors. The group developed malware capable of escaping security defenses by automatically detecting and rewriting code when flagged.

Anthropic identified new categories of threat actors misusing Claude, including those developing software for weapons design, including firearms, missiles, armed drones, bombs, and other munitions, as well as targeting and controlling systems for such weapons. The report detailed incidents involving affiliates of the ShinyHunters cybercrime collective, linked to attacks on major corporations worldwide. Jacob Klein, head of threat intelligence at Anthropic, stated that AI models have become more capable, raising new risks.