Anthropic report reveals how AI was misused in missile development

Anthropic reportedly uncovered a Yemeni weapons group that used its Claude AI model in an attempt to develop missile guidance software and foster the development of missile and rocket systems underscoring rising concerns over the use of advanced AI by non-state actors in military programs.

The unidentified factions were engaged in three arms programs including a guided rocket, a ballistic missile with a range exceeding more than 2,000 kilometers and a missile family known as R2000 the Anthropic stated in its latest report on detecting and countering AI misuse today.

The Yemen-based group used Claude AI to develop guidance, navigation, and automating away software engineer tasks.

Anthropic stated the group that it exposed in Yemen conducted a live rocket test and later consulted Claude to analyze a suspected launch failure. The group also created a local simulation site environment enabling partial weapon development to continue without ongoing access to Claude or standard engineering software environments.

Furthermore, Anthropic said it suspended accounts associated with the action, shared intelligence with partners and enhanced compliance measures following the inquiry.