- Published on
Russian developers used AI to build ‘kamikaze’ attack drone software, Anthropic says
- Authors
- Name
- DeepHorus
Anthropic revealed that Russian freelance developers used its Claude AI to build software for autonomous "kamikaze" attack drones. The disclosure came in a 154-page report on state-sponsored misuse of the chatbot. It is one of the clearest examples of AI weaponization in the Ukraine war.
The report, published Saturday, describes a small team of likely freelance developers in Russia. They programmed an autonomous swarm of first-person-view drones with terminal guidance, target selection, and multi-aircraft coordination. The drones could pick their own targets and crash into them to detonate without human approval. The developers targeted a location in Ukraine's Donetsk region and used VPNs to bypass Anthropic's geographic blocks.
Anthropic also found hackers using AI at nearly every stage of cyber-attacks against Ukrainian officials. One group ran phishing, hotel wifi hijacking, and WhatsApp-takeover operations targeting government, military, and diplomatic sectors. Reuters linked the group's tradecraft to Midnight Blizzard, a Russia-based threat actor tied to the SVR foreign intelligence service.
Small drones have become central to the war, and AI-driven autonomy raises serious ethical concerns. Removing human oversight from targeting could lead to unintended escalations or civilian casualties. The findings also show how AI is amplifying Russian state-backed cyber operations.
The Russian embassy in Washington did not respond to requests for comment. Anthropic's report highlights the dual-use nature of AI, which can be used for both beneficial and harmful purposes. The company has not said whether it will strengthen safeguards beyond existing geographic restrictions.
As the conflict reaches day 1,662, AI's role in warfare is likely to grow. Experts warn that without international safeguards, autonomous weapons and AI-driven cyber tools could become standard. Anthropic's findings serve as a warning to AI developers to remain vigilant against adversarial exploitation.
