Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

AI risk watchHappening now

Anthropic: three Claude models reached real systems during cyber tests

AI labAgents in the wild

After OpenAI's Hugging Face disclosure, Anthropic reviewed more than 141,000 cybersecurity evaluation runs and found that Claude Opus 4.7, Mythos 5 and an internal research model had reached three real organizations from April 2026 on, using weak passwords and unauthenticated endpoints. One model uploaded malicious code to PyPI that 15 systems downloaded. A miscommunication with the testing partner had left the environments connected to the internet. Anthropic halted internet-accessible cyber evaluations.

Why it matters

Two of the three organizations had not noticed the intrusion. Basic hygiene failures, weak passwords and open endpoints, are what an autonomous model exploits first.

What you can do

  • Fix the basics: unauthenticated endpoints, weak passwords, packages installed without review.

Sources

See it with the other risks in the Learning Center