Russian Hackers Use Claude to Auto-Rebuild Malware After Detection — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic attributed the campaign to group GTG-20006, aligned with Midnight Blizzard (APT29/Cozy Bear), whose AI agents autonomously modify and rebuild malware whenever monitoring agents detect a security product has flagged the artifacts.
- The GTG-20006 toolkit included two Windows implants, a mobile exploitation kit, a browser credential stealer, a phishing platform mimicking government targets, and an administrative console for managing compromised accounts.
- The actor targeted more than 20 organizations — government ministries, defense and intelligence bodies, embassies, think tanks, and defense-industrial firms — primarily in Ukraine and Europe, with additional victims in the Middle East and maritime agencies in Asia.
- The campaign overlapped with the CaptiveCrunch operation documented in July–August 2026 by ReliaQuest, Microsoft, Google, and Lumen Black Lotus Labs, including a sub-operation that compromised three hospitality vendors and used DNS hijacking on hotel guest Wi-Fi.
- Delivery implants included Windows malware (PowerChrome, WUEngine, Shadow C2, MiniPlasma, CloudSyncSvc), Android surveillance tool GiftDrop (rebranded from GiftsExpress RAT), and iOS payload DarkSword, paired with headless-browser WhatsApp takeovers that bulk-exported Russian and Ukrainian conversations.
- In a separate intrusion, the actor abused VPN appliance credentials to hijack a North African government technology authority, exfiltrating over 300,000 national identity records and commercial registry data for more than half a million companies.
- Anthropic concluded that AI has 'inverted the cost back onto defenders' — what previously slowed an attacker's operational tempo (deploying a new detection) no longer does when the attacker can autonomously mutate around it.
- Beyond malware development, the actor used AI to register phishing domains, set up hosting infrastructure, send lures, and monitor command-and-control channels for successful compromises across the entire operation.
Why it matters: Anthropic is naming a specific Russia-linked espionage group (APT29/Midnight Blizzard) as the first major case where AI doesn't just assist hackers but autonomously rewrites their malware in response to defenders' detections — the 300,000 leaked national identity records and 20+ compromised government/defense targets show the operational reach, and the shift means signature-based detection alone is now insufficient against state actors with AI in the loop.
Ask SkimNews
