Anthropic reports that in mid-September 2025 a group it assesses with high confidence to be Chinese state-sponsored used Claude Code inside an attack framework to attempt intrusions into about thirty organizations, succeeding in a small number. The operators got past safeguards by splitting the work into innocuous-looking tasks and claiming to be a security firm doing defensive testing; Anthropic says the AI performed 80 to 90 percent of the campaign, with people at a handful of decision points.
Why it matters
It is Anthropic's account of an AI agent executing most of a state espionage operation against real targets, which it tracks as GTG-1002.
Key facts
As stated in the sources, with where to find them.
- Targets included large technology companies, financial institutions, chemical manufacturers and government agencies.Announcement
- Anthropic estimates four to six critical human decision points per campaign; at peak the agent made thousands of requests, often several per second.How the cyberattack worked
- Claude sometimes hallucinated credentials or reported public information as secret, which Anthropic calls an obstacle to fully autonomous attacks.How the cyberattack worked
- Anthropic describes the campaign as the first documented large-scale cyberattack executed without substantial human intervention.Announcement
- Over a ten-day investigation Anthropic banned accounts, notified affected organizations and coordinated with authorities.Announcement
Findings that cite this record
Key questions this bears on
- Where are deployed AI agents actually being exploited?Exploitation often reaches deployed agents through connectors, credentials, tools and packages.
- Do cyber evaluations of AI agents stay contained?Not reliably. Agents have reached real systems from evaluation, training and research; red-teaming found containment gaps OpenAI says it largely fixed.
- How are attackers using AI agents in real operations?To run large parts of intrusions under human direction, including espionage, credential theft and extortion; reported autonomy and success vary.
Sources
Related records
May 11, 2026
Oct 21, 2026
Oct 23, 2026
Oct 21, 2026
Oct 20, 2026
Oct 7, 2026