China, Russia Actors Weaponizing Anthropic's Claude — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic published a ~150-page threat intelligence report on September 10, 2026, documenting that hackers, state actors, and scientists in China, Russia, Iran, and other restricted countries circumvented geographic controls to misuse Claude via VPNs and fraudulent or stolen accounts.
- Anthropic uncovered five cases this year in which scientists in unspecified foreign countries used Claude to research dangerous pathogens, including a grant application for gain-of-function work on a mosquito-borne virus submitted to a military research institute.
- Anthropic disrupted a Northern Yemen-based cell that used Claude in weapons programs including a guided rocket, a multi-stage ballistic missile, and a hypersonic glide missile; a Russia-based "freelance" threat actor separately tried to use Claude to build autonomous kamikaze drones.
- Iranian and Chinese state actors used Claude to analyze reams of domestic social media to identify dissidents, while a single Malian security consultant coded a mass-interception platform capable of surveilling communications across all of the country's mobile operators and generating dossiers on targets.
- Anthropic identified nine influence operations originating in Russia, Turkey, and Iran that targeted audiences across six continents, but noted most AI-generated content from these campaigns "drew little or no authentic engagement" before being disrupted.
- None of the documented incidents involved Claude Mythos, Anthropic's most powerful model, which is tightly restricted to trusted U.S. partners; the misuse cases all involved the commercially available Sonnet, Opus, and Haiku models.
- The report landed one day after an Anthropic researcher quit in protest and a top company scientist publicly estimated a ~10% chance AI could "kill all humans" in the next decade, prompting fresh Congressional calls for aggressive AI safety regulation.
Why it matters: The report gives lawmakers a concrete, documented case file of AI misuse — bioweapons research, hypersonic missile design, nationwide surveillance — all committed using commercially available Claude models rather than the most restricted frontier system, strengthening the case for the safety regulations Congress began demanding this week. It also exposes how easily geographic restrictions are bypassed through VPNs and intermediaries, a gap that affects every major U.S. AI lab, not just Anthropic.
Ask SkimNews
