Anthropic's Cyber Mission Targets the Verification Gap

Anthropic’s new Cyber Mission rests on a single claim: finding software vulnerabilities is now easy, and verifying, prioritizing and fixing them is the part that still takes months. The program is built around that gap.

It starts with two efforts. The Critical Infrastructure Defense Program gives trusted security providers access to frontier Claude models, on-site engineers and Anthropic’s threat research. The target is operational technology — power grids, water systems, transportation networks — that often cannot be taken offline for patching. Founding partners include Accenture, Booz Allen, CrowdStrike, Deloitte, Dragos, Hitachi, Nozomi Networks, Palo Alto Networks, PwC and Rockwell Automation.

The second effort, OSS Scanner, is an opt-in service for open-source projects, inspired by Google’s OSS-Fuzz. Enrolled projects get periodic free scans, and each report includes a proof of concept, an explanation and a suggested fix. Anthropic expects a true-positive rate above 90 percent. Funding also goes to the Python Software Foundation, Alpha-Omega and OpenSSF, and the Apache Software Foundation.

The volume is already there. In its Cyber Verification Program announcement, Anthropic said Project Glasswing partners found at least 129,000 verified vulnerabilities between April and July 2026, more than 33,000 of them rated critical or high severity. Anthropic now forecasts that AI will favor defenders within two years, though not necessarily in the near term.

The partner list shows where the work lands: integrators and operational-technology specialists, not the model lab, carry the verification and remediation. For any company buying AI-driven security services, the question to ask a vendor is how many findings they close, not how many they report. Source: Anthropic’s Cyber Mission announcement.