AI Misuse for Weapons Development by Houthi Rebels in Yemen

• •Houthi rebels in Yemen attempted to use AI for advanced missile development.
• •Anthropic's Claude AI was misused for guidance software and weapon design.
• •The actors built an offline toolkit despite account bans, indicating ongoing capabilities.

Anthropic's report reveals that users in northern Yemen, controlled by Houthi rebels, attempted to develop advanced missiles using the Claude AI model. The report, spanning eight months, indicates that these actors were involved in three weapons programs, including a hypersonic glide missile. Although the users did not succeed in operationalizing their devices, they conducted a failed test of a guided rocket and built an offline simulation toolkit. The report highlights the Houthis' ongoing efforts to enhance their military capabilities amid their conflict with Saudi Arabia. Anthropic blocked the accounts after identifying the misuse, but the actors had already made significant progress in their attempts. The report also mentions state-aligned influence operations in Iran, utilizing Claude for propaganda efforts. This incident raises concerns about the implications of AI in warfare, particularly in conflict zones.