Google opens Fairwind cyber defense program to trusted partners

Google has launched the Fairwind Program, a limited-access initiative for governments, Google Cloud customers, and cybersecurity partners. In its Google DeepMind Blog announcement dated September 2, 2026, the company says the program will provide advanced AI capabilities for proactively addressing cyber risks.

The first offering combines Gemini 3.8 Flash Cyber with Google’s CodeMender harness. The system is designed to identify vulnerabilities, reason through code changes, validate fixes, and produce patches ready for deployment. Google says work that could take security teams weeks to complete manually may be reduced to minutes inside an organization’s secure cloud environment, at a lower operating cost than traditional frontier models.

Who gets access

Initial participants include government and national cyber authorities, operators of critical infrastructure such as healthcare, telecommunications, energy, and finance, and technology platforms whose software supports large numbers of downstream users. More than 650 partners globally are already participating, according to Google.

The program is not an unrestricted release. Organizations must limit use to internal cybersecurity, incident-response, or penetration-testing teams and apply safeguards including multi-factor authentication. Google says access will expand over time while it works with industry, governments, and open-weight model communities.

For builders, the announcement highlights a practical shift from AI-assisted vulnerability discovery toward agentic remediation: tools that can propose and verify code fixes rather than simply report problems. Google Cloud customers outside Fairwind can still use CodeMender with publicly available models through the Gemini Enterprise Agent Platform, alongside Google’s AI Threat Defense offerings.

Source: Google DeepMind Blog