Anthropic reports Yemeni rebels attempted to use Claude AI to build guided weapons
Anthropic, an AI company, released a report Thursday revealing that a group of militants in Yemen attempted to use its Claude chatbot to develop guidance software for three sophisticated missile projects between December 2025 and August 2026. According to the report, the activities included efforts to build a guided rocket, a multi-stage ballistic missile with a range exceeding 2,000 kilometers, and a hypersonic glide vehicle variant. Anthropic stated that the group attempted a test firing of a guided rocket which appears to have failed. The company said it blocked the accounts involved after identifying the activity and passed information to government partners. Though mountainous northern Yemen is controlled by Iran-backed Houthi rebels, Anthropic did not explicitly identify the actors. The company emphasized that no evidence suggests the group succeeded in fielding an operational weapon. The incident highlights growing concerns about powerful AI models being available to non-state actors for weapons development, particularly as advanced AI reduces the technical expertise required for such projects.
Key facts
- Activity occurred between December 2025 and August 2026, involved Claude Haiku, Sonnet and Opus models
- Three weapons programs: guided rocket with homing guidance, multi-stage ballistic missile with 2,000+ kilometer range, multi-variant missile with hypersonic glide vehicle variant
- Test-fired guided rocket appeared to fail; users returned to Claude within hours to diagnose failure
- Anthropic blocked accounts after detecting activity and shared threat information with government partners
- No evidence suggests a functioning operational weapon was successfully developed
- Group had built offline simulation toolkit that survived the account ban
How different outlets framed it
Left-leaning coverage: Left-leaning outlets and progressive voices emphasize AI safety concerns and call for stronger government regulation. Senator Bernie Sanders stated AI threatens the economy and democracy, urging Congress to hold tech firms accountable. Some progressives highlight the broader pattern of AI misuse across biological weapons research and influence operations.
Right-leaning coverage: Right-leaning outlets emphasize the national security threat posed by Iran-backed Houthi rebels acquiring advanced weapons capabilities through AI, situating this within broader concerns about Houthi attacks on U.S. interests, Saudi Arabia, and Israel in the Middle East.
Limits of this briefing
Anthropic did not explicitly identify the actors as Houthis, though geographic location and operational details strongly suggest this. The company's decision to release the report amid broader AI safety debates may reflect both genuine security concerns and efforts to demonstrate the risks of powerful AI models as the company advocates for industry-wide development slowdowns.
Sources
- The Washington Post: Rebels used Anthropic's AI bot to develop guided weapons, report says Centre
- Al Jazeera: Anthropic claims Claude AI used for missile projects, global espionage Centre
- SecurityWeek: Users in Houthi-Held Yemen Tried to Develop Advanced Weapons With AI, Anthropic Says Centre
- RedState: New: Anthropic Foils Houthis' Bid to Build Missiles With Claude AI — For Now Right
- Rolling Out: Anthropic's safety case faces its own scrutiny Left