Google launches Fairwind to protect AI systems

Google has launched Fairwind, a program that allows governments, large companies, and cybersecurity partners to use AI to find and fix software vulnerabilities before attackers can exploit them.

Initial access will be limited to a selected group of organizations. According to Google, more than 650 partners worldwide are already participating, including public agencies, critical infrastructure operators, and technology platforms.

AI to detect and fix vulnerabilities

Fairwind combines the specialized Gemini 3.8 Flash Cyber model with CodeMender, a Google tool that analyzes code, proposes fixes, and checks that the changes work before they are deployed.

The difference from a system that only detects problems matters. Finding a vulnerability alerts you to a risk, but does not remove it. Google’s proposal is for AI to also generate a deployment-ready patch inside the organization’s secure environment.

Google says this process can reduce the time needed to fix certain vulnerabilities from weeks to minutes. The company also says the model delivers advanced capabilities at a lower operating cost than larger frontier models, although the announcement does not publish independent comparative results.

“The time between detecting a vulnerability and fixing it is the defender’s advantage,” Google says.

Who can use it

The program prioritizes organizations whose security affects many other people and services:

• Governments and national cybersecurity authorities, to protect public networks and citizen services.
• Critical infrastructure operators, such as hospitals, telecommunications networks, energy companies, and financial institutions.
• Technology platforms, which can strengthen software components used by millions of users and businesses.

Participating organizations will have to apply strict controls. Access will be restricted to internal cybersecurity, incident response, or penetration testing teams, and measures such as multifactor authentication will be mandatory.

What changes for businesses

Fairwind is not an open launch for every user. It is a controlled-access program for Google Cloud customers and partners considered strategic.

Even so, Google says any Google Cloud customer can use CodeMender with publicly available models through Gemini Enterprise Agent Platform. They can also combine it with the company’s AI security solutions.

In practice, the tool’s usefulness depends on how it is integrated into a company’s processes. AI can speed up review and patch creation, but human teams will still need to validate the changes, check that they do not break other systems, and decide when to deploy them.

More cybersecurity funding

The program is part of a broader Google initiative to strengthen cybersecurity. Through Google.org, the company says its global funding in this area has now exceeded $100 million.

Google also reports $36 million allocated to 35 cybersecurity clinics in the United States, which have provided free support to more than 1,250 hospitals, public school districts, and municipal services.

What remains to be seen is whether Fairwind can turn its promise of autonomous remediation into measurable results: less time to patch, fewer errors introduced, and real protection before attackers can exploit vulnerabilities.