Anthropic has said that Russian developers used artificial intelligence to create software for "kamikaze" attack drones, The Guardian reported. The claim appeared in the company's September 2026 report on detecting and countering misuse of its AI systems.
The same report, according to The Guardian, also describes other cases in which Anthropic says its technology was misused. The company said it blocked what may have been an attempt to use AI to make biological weapons, a case the BBC also covered. Separately, Anthropic said Iran used its American AI model to target U.S. Navy warships, according to The Wall Street Journal. TRT World reported that the Houthis secretly used Claude AI in an attempt to build guided missiles.
The disclosures place Anthropic among AI companies publicly documenting how their models are used in military and weapons-related work. The Guardian's briefing frames the drone software case within the wider Ukraine war, where both sides have relied heavily on unmanned aircraft. The report does not detail how the Russian developers accessed the system or what stage the drone software reached.
Anthropic's report covers several alleged misuse cases at once, spanning drone software, biological weapons and missile development. Each claim rests on the company's own account of activity on its systems.
What to watch next: whether Anthropic or other AI developers release further technical detail on these cases, and how governments respond to AI models being used in weapons development.
