Malware: Company, international 2025
A Chinese group that Anthropic assesses with high confidence to be state-sponsored (GTG-1002) broke attacks down into harmless-looking subtasks and posed as a security firm in order to misuse Claude Code.
- Incident · 09.2025 · worldwide (around 30 targets), international
- Evidence: confirmed by an authority
- Malware
- Data breach
- AI link
As of
What happened
The AI mapped systems, wrote exploit code, collected login data and exfiltrated data. The intrusions succeeded in a small number of cases; Anthropic blocked the accounts and informed those affected and the authorities.
Role of AI: According to Anthropic, the AI did around 80–90% of the work (reconnaissance, exploits, login data, data exfiltration); humans intervened only at a few decision points. The AI sometimes hallucinated login data.
- Affected
- Company (anonymous)
- Industry
- several industries
- Loss
- no reliable figure
- Business interruption
- unknown
- Ransom demanded / paid
- no / no
- Data affected
- true
- Bankruptcy
- no
- Entry point
- Vulnerabilities in the targets’ systems; Claude Code, via a jailbreak, as a largely autonomous attack tool
What would cyber insurance have paid?
That depends on the terms. Whether those affected were insured: unknown.
-
Immediate help in an emergency
Hotline, IT forensics, crisis PR and legal advice
Mostly covered 7 of 9 business products cover it fully, as an add-on or with limits.
-
Costs of getting data, software and systems running again
Mostly covered 6 of 9 business products cover it fully, as an add-on or with limits.
-
Third-party claims after a data breach
Liability for data protection and network security breaches
Mostly covered 7 of 9 business products cover it fully, as an add-on or with limits.
-
Proceedings before authorities
Costs of FDPIC or GDPR proceedings, usually not fines
Mostly limited 6 of 9 business products cover it fully, as an add-on or with limits.
Overview based on public policy terms, as of 06.11.2026. Only your policy and its general terms and conditions (GTC) are binding. No confirmation of cover.
What we can learn
AI drastically lowers the effort and increases the speed of attacks; patching and monitoring access become more urgent.
Sources
- Disrupting the first reported AI-orchestrated cyber espionage campaign (opens in a new window) Anthropic · 2025-11-13
«The attackers used AI’s “agentic” capabilities to an unprecedented degree»
- Chinese Hackers Use Anthropic's AI to Launch Automated Cyber Espionage Campaign (opens in a new window) The Hacker News · 2025-11