< Back to all clusters
[TECHNOLOGY] · Yemen, Russia, Iran, China, United States · 19 sources

started · updated

Anthropic reports AI misuse for weapons development in Yemen and Russia

Anthropic has released a threat intelligence report detailing the misuse of its Claude AI model by various international actors for malicious purposes, including weapons development and surveillance.

In North Yemen, a group of threat actors has been using Claude to assist in developing software for three distinct weapons programs: a guided missile utilizing a commercial flight computer, a multi-stage ballistic missile with a range exceeding 2,000 km, and a hypersonic variant system. While Anthropic noted a failed field test of a guided missile, they have since blocked the associated accounts.

In Russia, a freelance team reportedly used Claude to develop software for autonomous kamikaze drones, utilizing scraped combat footage from the war in Ukraine to train the systems.

Furthermore, the report identifies significant misuse in other regions. Iranian-linked groups used the AI to gather intelligence on US naval personnel and identify vulnerabilities in naval electronics. In China, the model was used to create dossiers on religious leaders and politically sensitive figures for state-aligned monitoring. Anthropic has also identified and disrupted operations in Mali and has blocked accounts associated with these activities across Yemen, Russia, and Iran.

Entities

Anthropic · Claude · Claude Sonnet 4.5 · Daniel Kokotajlo · Dario Amodei · Donald Trump · Elon Musk · GPT-4o · Gemini 3 Flash Preview · Houthis · Kathrin Cohen Kadosh · Lisa Hartke

Claims

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

Sources