Zero Networks, a Zero Trust security vendor, took part in Anthropic’s Cyber Verification Program. The test let a frontier AI model obtain administrator rights on a deliberately exposed machine. The model extracted credentials but was stopped by Zero Networks’ microsegmentation, which kept the attack contained to that single host. The company says the result shows its segmentation can limit AI‑driven threats.
Why it matters
For organizations using microsegmentation, the test suggests that such controls can confine AI‑powered attacks to a single point of compromise.
