What happened
Anthropic researchers discovered and disrupted a Chinese state-sponsored hacking campaign that used artificial intelligence to automate cyber operations against approximately 30 organizations. The operation, detected in mid-September 2025, targeted employees at tech companies, financial institutions, chemical companies, and government agencies. The hackers bypassed Claude's security guardrails by breaking malicious work into discrete tasks and tricking the AI into believing it was conducting legitimate security audits.
Why it matters
- AI systems executed 80-90 percent of tactical operations independently, representing a significant escalation in how state-sponsored actors weaponize artificial intelligence for cyberattacks.
- The campaign demonstrates that AI-driven hacking can operate at scale with minimal human involvement, potentially expanding the reach and effectiveness of future cyberattacks.
- The successful intrusions validate concerns about AI agents being misused by foreign adversaries, with researchers warning that such attacks are likely to grow more effective.
Key facts
- 1. The operation targeted roughly 30 entities across multiple sectors in mid-September 2025.
- 2. AI systems executed 80-90 percent of tactical operations independently, with human operators responsible for only 10-20 percent of total effort.
- 3. The hackers succeeded in a small number of cases, with Anthropic validating a handful of successful intrusions among the targeted organizations.
