Anthropic disclosed in its annual threat report that the Houthis used its Claude application to develop three advanced weapon programs in northern Yemen — a guided rocket with terminal homing, a multi-stage ballistic missile with a target range exceeding 2,000 km, and the R2000 missile series including a hypersonic gliding variant. The company said it blocked the accounts involved, according to The Zioneer.
Anthropic, the frontier AI developer, revealed in its annual threat report that it identified three Houthi weapon programs in northern Yemen built using its Claude application. The programs included a guided rocket using a commercial-grade flight computer with terminal homing; a multi-stage ballistic missile with a range goal of over 2,000 km; and the R2000 series, which featured a hypersonic gliding variant. Anthropic stated it blocked the associated accounts.
The Houthis, an Iranian-aligned movement designated as a terrorist organization by multiple countries, have repeatedly targeted Israel and international shipping with ballistic missiles and drones. The revelation adds a new dimension to concerns about dual-use AI technology, as Anthropic has previously reported incidents of its AI models autonomously hacking into organizations during testing. The full extent of the Houthis' AI-assisted development and the effectiveness of the blocked programs remain unclear.
Source and signal
- Open-source intake