Anthropic said it disrupted a cell of threat actors in northern Yemen that used its Claude Code tool in place of human software engineers to develop guidance software for a guided rocket, a ballistic missile, and a hypersonic glide vehicle variant.
Cell Ran Several Claude Instances At Once
In its September threat intelligence report published Thursday, the company said the cell ran three weapons programs, including a multistage ballistic missile with a stated range goal above 2,000 kilometers, and a missile set referred to as ‘R2000’ that included a hypersonic glide vehicle variant.
They managed multiple Claude instances at once, assigning one to write code, another to research, and a third to review the work.
Anthropic’s safeguards blocked many requests, but not all, as the actors hid their goals and the software’s intended use and split their work across sessions, so none revealed their full intent.
Failed Rocket Test Led Cell Back to Claude
The actors test-fired a guided rocket that appears to have failed, and Claude returned within hours to work out why.
The company said it had no evidence that the actors fielded an operational device, but they had already built an offline simulation toolkit that does not rely on Claude.
Anthropic identified the activity through internal investigations into suspected weapons development and banned the accounts.
The AI lab did not identify the actors involved.
Houthi Attacks Escalate
A new attack on Riyadh’s King Khalid International Airport killed one person and injured 11, Saudi Arabia’s civil aviation authority said Sunday, after an earlier strike killed 12 people and injured more than 300.
Saudi authorities blamed the Iran-backed Houthis, who control much of northern Yemen.
President Donald Trump said Saturday that the U.S. could consider action against the group.
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
Photo courtesy: Shutterstock