# Zero Networks Reports Microsegmentation Blocks AI Attack

Zero Networks reports that, during an Anthropic Cyber Verification Program test, an autonomous Claude model compromised an intentionally exposed machine but could not move laterally into systems protected by microsegmentation. The result is a company-reported test finding, not a government procurement action: no solicitation, contract award, funding amount, or procurement contact was identified. For government buyers and contractors, it offers a concrete example to consider when assessing or demonstrating network segmentation and zero-trust controls against automated attacks.

- Security teams and contractors can use the reported scenario to evaluate whether their segmentation controls limit lateral movement after an initial compromise.
- The result may inform discussions about security capabilities in contractor offerings, but it does not establish a new government requirement or an active procurement opportunity.
- The reported findings come from a participating company; procurement professionals should distinguish this test result from independently established performance evidence.

**Jurisdictions:** international
**Industries:** Information Technology
**Topics:** Cybersecurity
**Published:** November 05, 2026

### Government Entities
- Cybersecurity and Infrastructure Security Agency (CISA)
- National Cyber Security Centre (NCSC)

### Vendors
- Zero Networks (Evaluation participant)
- Anthropic (AI model developer and Cyber Verification Program organizer)

### Key Quotes
> Segmentation is one of the few controls that changes that math regardless of how the attacker thinks or how fast it moves, and this test is a concrete demonstration of why that matters.
> — Dr. Chase Cunningham, Cyber Security Expert specialising in zero trust architecture

### Sources
- [AI attacker blocked by microsegmentation in Anthropic test](https://securitybrief.co.uk/story/ai-attacker-blocked-by-microsegmentation-in-anthropic-test) - SecurityBrief UK