Zero Networks, leading provider of Zero Trust security solutions, has announced results from participation in Anthropic Cyber Verification Program showing frontier AI model unable to move beyond intentionally exposed machine after gaining administrator access and extracting credentials. In evaluation Zero Networks microsegmentation prevented model from breaching protected systems and identities containing attack to initial point of compromise.
Frontier models increasingly capable of discovering vulnerabilities obtaining credentials reasoning through multi-step attack paths operating with less human intervention that makes traditional question whether attacker can compromise endpoint only part of problem more consequential what attacker can reach after initial compromise Zero Networks conducted evaluation to test defensive side if highly capable AI attacker gets foothold and acquires credentials can architecture prevent turning foothold into lateral movement evaluation showed answer was yes.
"AI is going to make attackers faster, more persistent and increasingly autonomous. Security teams aren't going to win by trying to out-click or out-think machines operating at machine speed," said Benny Lakunishok, CEO and co-founder of Zero Networks. "The answer is to build environments where compromising one machine doesn't give the attacker somewhere else to go. That's what this test demonstrated: the attacker got in, had credentials and had every incentive to keep moving. The architecture stopped it."
Methodology Zero Networks built large lab environment designed to resemble production network including Windows Linux macOS systems Kubernetes heterogeneous workloads then applied Zero Networks microsegmentation to protect assets identities within environment one machine intentionally left unsegmented to serve as control and initial point of compromise frontier AI model deployed as autonomous attacker with administrator access no restrictions on how could attempt to attack environment tested ability to compromise systems obtain credentials pursue lateral movement using offensive security tools techniques Claude attempted 18 distinct attack paths using 23 offensive tools and 22 MITRE ATT&CK techniques including Kerberoasting Active Directory credential extraction and Pass-the-Hash attempt against primary target.
"What OpenAI disclosed should worry people well beyond the security team: autonomous agents finding each other, colluding, and escalating access without a human in the loop," said Dr. Chase Cunningham, cybersecurity expert specializing in zero trust architecture. "The infrastructure the world actually runs on was never built to assume the attacker on the other end doesn't get tired. 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."
About Zero Networks
Zero Networks makes microsegmentation easy automated achievable helping organizations stop lateral movement keep business running even when attackers get inside automatically creating enforcing identity-based least-privilege policies across users workloads AI agents using controls already built into environment with no network redesign more than 400 enterprises rely.
