Ukraine desk brief

Anthropic reports Russian developers used AI to create autonomous ‘kamikaze’ drone software targeting Donetsk

Anthropic’s report alleges Russian-linked hackers used AI to build attack drones and cyber operations targeting Ukrainian government and military, though independent verification is pending.

What happened

Anthropic, the AI company behind the chatbot Claude, reports that a small team of likely freelance developers in Russia used its AI system to program software for an autonomous swarm of kamikaze attack drones. These drones were designed for self-targeting and terminal strikes without human intervention, with a persistent focus on locations in Ukraine’s Donetsk region, according to a 154-page report released by Anthropic.

Outside Brief is treating this as a source-led account. Any disputed responsibility, casualty figure, battlefield claim or single-source assertion should be treated as unconfirmed/hearsay unless confirmed by another reliable source or a named official. The report also details how hackers, allegedly linked to Russia-based threat actor Midnight Blizzard, deployed AI at multiple stages of cyberattacks targeting Ukrainian government, military, and diplomatic sectors. These operations reportedly included phishing, hijacking hotel Wi-Fi networks, and WhatsApp account takeovers, with AI assisting in malware evasion by automatically modifying flagged code until it slipped past security defenses.

Anthropic's findings come amid increased physical attacks on Ukrainian infrastructure, including recent jet-powered drone strikes on petrol stations in Kyiv that have caused casualties and disrupted fuel supplies. While the report links the AI-assisted drone software development to Russia, independent verification of direct Russian state involvement remains outstanding.

Known from the source

  • Anthropic released a detailed 154-page report on the misuse of its Claude AI system.
  • The report alleges Russian freelance developers used the AI to build autonomous kamikaze drone software targeting Ukraine’s Donetsk region.
  • Anthropic identified AI-assisted cyberattacks targeting Ukrainian government, military, and diplomatic sectors.
  • The hacking group’s tradecraft aligns with Russia-linked threat actor Midnight Blizzard.
  • Kyiv has recently suffered jet-powered drone attacks on petrol stations, causing deaths and injuries.

What remains unclear

Russian President Vladimir Putin continues to attribute the conflict to Western policies and alleged Western attempts to integrate Ukraine into NATO, framing these as causes of the ongoing hostilities. This context underlines the broader geopolitical tensions that accompany the technological developments described in Anthropic's report.

What remains unclear: Confirm whether the central claim is corroborated; until then treat it as unconfirmed/hearsay. Verification of autonomous drone software development and deployment by Russia or its proxies. Independent confirmation of the AI-assisted cyberattacks and direct links to Midnight Blizzard or Russian intelligence. Corroboration of casualty and damage reports from recent drone strikes in Kyiv.

Evidence note

This story contains report-led claims. The article keeps those claims attributed and treats them as unconfirmed/hearsay unless independently corroborated.

Original source: The Guardian Ukraine. Open the source.

Outside Brief note: this story keeps the main source visible and separates what is reported from what remains unclear.