On September 2, 2026, Google announced the Fairwind Program, which it describes as a limited access program for governments and trusted partners to use its most advanced cyber defense capabilities. Participation is drawn from government agencies, critical infrastructure operators, and core technology platforms, with more than 650 global partners named, and access is restricted to employees inside a participant’s own cybersecurity, incident response, or penetration testing teams.
The technical payload is two things working together: the Gemini 3.8 Flash Cyber model and the CodeMender harness. Google’s claim is that the pair lets defenders find, verify, and fix vulnerabilities at agentic scale, generating verified, deployment-ready patches in minutes inside a customer’s own cloud, in place of manual remediation that takes weeks. The word doing the work there is “verified” - the difference between a model that proposes a diff and a harness that proves the diff fixes the bug without breaking the build is the whole difference between a demo and a tool a national infrastructure operator can run.
Google paired the launch with philanthropic numbers: more than 100 million dollars in total global cybersecurity funding through Google.org, including 36 million dollars for 35 cyber clinics that have supported over 1,250 hospitals, public school districts, and municipal utilities in the United States.
The strategic reading is that the offense-defense balance in agentic security is now being contested deliberately rather than left to chance. The same capability that lets an agent find and exploit a flaw at scale lets a defender find and patch it at scale, and the deciding variable is who gets the good harness first and under what controls. Gating the strongest cyber model behind a vetted-partner program, with access limited to named security teams, is Google’s answer to that question, and it is the model other labs will be measured against.