< Back to all clusters
[TECHNOLOGY] · China, Iran, Yemen, Russia, United States · 37 sources

started · updated

Anthropic reports Claude AI misuse for weapons and espionage

Anthropic has released a comprehensive threat intelligence report detailing the widespread misuse of its Claude AI models for malicious activities. The report identifies several key areas of concern, including cyber espionage, military development, and biological research.

In the realm of military and engineering, actors linked to Russia, China, and the Houthi rebels in Yemen have reportedly used Claude and Claude Code to assist in weapons development. Specifically, Houthi-controlled groups attempted to develop guidance, navigation, and control software for missiles, including a multi-stage ballistic missile and a hypersonic glide vehicle variant. A China-based actor also used the model to simulate electronic warfare and air defense suppression involving Taiwan.

Regarding intelligence and surveillance, Iranian-linked operatives utilized Claude to gather targeting information on US Navy assets in the Middle East by analyzing open-source data, including military photographs and transponder information. Additionally, the report notes that state-sponsored actors from China and Iran have used the AI for surveillance and influence operations targeting political dissidents and diaspora communities.

Anthropic also identified attempts to use the technology for high-risk biological research involving pathogens like smallpox and bird flu. Furthermore, the company reported significant efforts by Chinese AI labs, such as Alibaba and DeepSeek, to conduct illicit distillation attacks to extract Claude’s reasoning capabilities. Anthropic has responded by banning offending accounts and enhancing its security guardrails.

Entities

Alibaba · Anthropic · China · Claude · DeepSeek · Houthi rebels · Houthis · METR

Claims

What the coverage asserts, and how many sources carry each claim.

Sources

about 11 hours ago
about 12 hours ago