AI Chatbot Accidentally Hired by International Freelance Drone Hit Squad

KEY POINTS

•Anthropic found Russian freelancers from a regional university using Claude AI in mid-2026 to develop drone swarm software for autonomous strikes in Ukraine.
•Chinese threat actors used Claude to generate 200-page anti-torpedo weapons proposals and analyze US naval systems through role-playing tactics.
•Anthropic also uncovered Yemen-based rocket engineers and Iran-linked actors using Claude for missile control software and naval targeting data, banning all involved accounts.

Anthropic’s 2026 blockbuster report reveals Russians from a regional university—because who needs state backing?—used the AI chatbot Claude to write autonomous drone swarm code that picks, attacks, and then heads home like it just beat rush hour traffic. From May 2026, these nine sneaky freelancing Russians fooled AI’town by VPN gymnastics, testing software on real hardware and providing a Donetsk coordinate for target practice. Meanwhile in China, Claude crafted a 200+ page anti-torpedo naval proposal polished by role-playing hostile reviewers (AI critiquing AI, the future of passive-aggressive meetings!). Up north in Yemen, rocket engineers relied on Claude to fix a failed test-fire rocket, apparently being AI’s QA testers. An Iran-linked group used Claude for US naval stalking homework, scraping captions to ID sailors. Despite bans and VPN tricks, one tech whistleblower resigned fearing AI-triggered apocalypse by decade’s end. We’re not saying Claude’s making drone armies, just that it’d probably drone on at meetings too.