US artificial intelligence company Anthropic has revealed that Russian-linked actors used its Claude AI model to conduct cyber-espionage against Ukrainian and European government targets, run influence operations in Africa and Moldova, and develop autonomous targeting software for kamikaze drone swarms. A hacking group tied to Russia's SVR foreign intelligence service — known in cybersecurity circles as Midnight Blizzard — used Claude to attack over 20 organisations, including defence ministries, embassies and intelligence agencies, automating tasks from writing malware to extracting data from compromised networks. Separately, Russian-linked developers used Anthropic's Claude Code tool to build an autonomous targeting system for FPV drone swarms, training AI models on combat footage from Ukraine and designating a location in the Donetsk region as a test strike target; Anthropic said it had blocked all identified misuse, strengthened its security systems, and shared findings with government and industry partners.