# Anthropic says attackers now use Claude to run whole cyberattacks

> Anthropic banned accounts using Claude to automate espionage, surveillance and model copying.

*A September threat report details Russia- and China-linked campaigns and accuses seven Chinese labs of harvesting Claude to train rivals.*

By Behzad Hosseini · FeaturedDaily
Canonical: https://featureddaily.com/news/anthropic-says-attackers-now-use-claude-to-run-whole-cyberattacks

**What happened:** Anthropic published its fourth threat intelligence report on 10 September 2026, detailing misuse of Claude between December 2025 and August 2026.

**The numbers:** The report covers six areas: cyber operations, influence operations, surveillance, scams, weapons-related misuse and illicit model distillation. Nine separate influence campaigns were logged, alongside espionage activity hitting more than 20 organisations.

**Why it matters:** Attackers are no longer just asking Claude for advice or code. Anthropic says they're using it to orchestrate entire attacks, from reconnaissance to malware rewrites, across multiple stages.

**The context:** A Russia-linked group (GTG-20006, consistent with Midnight Blizzard) used Claude to automate reconnaissance and malware development against drone makers and Ukrainian officials, according to [Anthropic's report](https://www.anthropic.com/threat-intelligence-report-september-2026). Its agents reportedly watched security products and rewrote malware in a loop until it evaded detection.

**Who's affected:** A China-based group (GTG-14010) used Claude to convert chatter from over 100 monitored WhatsApp groups and dozens of Telegram channels into structured data tracking Uyghurs in Syria, and to draft restricted briefings on dissidents and diaspora communities, Anthropic said.

**The catch:** Anthropic accused seven Chinese labs, Alibaba, DeepSeek, Moonshot, Xiaomi, Zhipu, SenseTime and MiniMax, of using thousands of fraudulent accounts to harvest Claude's responses for training their own models, a practice known as distillation.

**The rivals:** Other cases included the ShinyHunters criminal group using Claude to speed up system takeovers, and two Chinese undergraduates running an automated exploit "foundry", according to [Reuters](https://www.usnews.com/news/world/articles/2026-09-10/anthropic-disrupts-russian-chinese-ai-campaigns-targeting-its-claude-models). Most cases involved older Claude Haiku, Sonnet and Opus models rather than Fable or Mythos-class systems.

**What's next:** Anthropic said it has banned all accounts identified in the report.

## Key takeaways

- Russia-linked hackers used Claude to rebuild malware after detection
- China-based group tracked Uyghurs via 100+ WhatsApp groups
- Anthropic accuses seven Chinese labs of fraudulently copying Claude

## Sources

- [Countering misuse of AI: September 2026](https://www.anthropic.com/threat-intelligence-report-september-2026) — Anthropic, 2026-09-10
- [Anthropic disrupts Russian, Chinese AI campaigns targeting its Claude models](https://www.usnews.com/news/world/articles/2026-09-10/anthropic-disrupts-russian-chinese-ai-campaigns-targeting-its-claude-models) — Reuters via US News, 2026-09-10