Russian Developers Used Claude AI to Build Kamikaze Drone Software

Russian teams exploit AI for autonomous weapons development

A small team of likely freelance developers in Russia used Anthropic's Claude AI chatbot to build software for autonomous kamikaze drone swarms capable of selecting and attacking targets without human intervention, according to a 154-page report released by the AI company.

The developers programmed the drones to handle terminal guidance, target selection, and coordination between multiple aircraft. During development, the team repeatedly selected locations in Ukraine's Donetsk region as test targets and used VPNs to bypass Anthropic's geographic restrictions. The drones were designed to identify targets independently and crash into them to detonate.

AI-powered cyberattacks target Ukrainian officials

Anthropicโ€™s investigation also uncovered cybercriminals and state-backed hackers using AI throughout nearly every stage of sophisticated cyberattacks. One hacking group conducted phishing campaigns, hotel WiFi hijacking, and WhatsApp takeover operations against Ukrainian government, military, and diplomatic personnel.

The group's methods aligned with Midnight Blizzard, a Russia-based threat actor previously linked by the U.S. government to Russia's SVR foreign intelligence service, according to Reuters. The hackers allegedly built an AI system that automatically detected when security defenses flagged their malware and rewrote the code until it evaded detection. The Russian embassy in Washington did not immediately respond to requests for comment.

Escalating attacks on Kyiv infrastructure

The disclosure comes amid intensifying Russian drone strikes on Ukraine's capital. Jet-powered drones hit at least two Kyiv petrol stations during Friday morning rush hour, killing two people and injuring 12, according to Mayor Vitali Klitschko. The attacks followed a similar strike on Thursday, with all three targeting Ukrnafta, a state-owned oil and gas company.

A list of 18 Kyiv petrol stations expected to be targeted by Russian drones began circulating on Telegram Wednesday. The three stations struck were on that list. Russia has conducted nearly constant attacks with faster jet-powered drones for two weeks, killing multiple Kyiv residents and disrupting daily life in the capital.

Why it matters

The Anthropic report provides concrete evidence that AI systems are being weaponized in active conflicts, not as a theoretical future threat but as operational tools today. For AI companies, it demonstrates the difficulty of preventing misuse despite geographic blocks and usage policies. For military and security leaders, it signals that adversaries are rapidly integrating AI into both kinetic weapons and cyber operations, potentially outpacing defensive countermeasures. The autonomous targeting capability described raises urgent questions about accountability and escalation risks when machines make lethal decisions.

Diplomatic developments

Russian President Vladimir Putin blamed Western pressure and attempts to bring Ukraine into NATO for the conflict, speaking before the BRICS summit in India. A Russian official will attend a G20 energy meeting in Houston next week, following controversy last month when Russian Finance Minister Anton Siluanov attended a G20 meeting in North Carolina—his first in-person appearance at the forum since the 2022 invasion.

The U.S. House of Representatives is expected to consider a Russian sanctions package next week, though passage remains uncertain despite Senate support. Democratic lawmakers have expressed concern the bill expands President Donald Trump's tariff authority in ways that could affect U.S. allies.

These details were first reported by The Guardian.