Keeping defenders one step ahead
Using frontier AI capabilities can help people to rapidly discover – and fix – critical vulnerabilities. But in the wrong hands, the same capabilities can become an equally powerful threat.
The Fairwind Program gives high-priority defenders (like governments, healthcare providers, and telecommunications services) early access to advanced models that help them build better defenses, before new threats arrive. So defenders have an early advantage, to help them protect vital infrastructure – which in turn protects people who rely on those systems.
Trusted defenders and industry leaders are already putting Gemini 3.8 Flash Cyber into practice:
“Armadin is trusted to defend the world’s largest companies and organizations by running swarms of agents to find and remediate exploitable risks. In our evaluations, Gemini 3.8 Flash Cyber matched larger frontier models at a lower price and with best-in-class latency. This model is a critical tool in the defender’s arsenal and extends the Pareto frontier for cyber work.”
David Slater, Founder & Chief Architect, Armadin
“To outpace automated threats, defenders need model performance that operates at the speed of the attack. Early testing of Gemini 3.8 Flash Cyber proved the model to be exceptionally fast, and it performed above its model class across a number of cybersecurity tasks, including vulnerability hunting.”
Charlie Sestito, Director, Office of the CTO, Palo Alto Networks
“We pointed Gemini Cyber at our public repos where the stakes are higher. Over the course of our two day-trial, it held its own against much larger engines on critical and high severity findings, and it cut the triage noise our teams work through. It did that at Flash-style cost, cheap enough to run continuously rather than in occasional sweeps.”
Mayank Upadhyay, CSTO, Snowflake
“In our evaluations targeted at web exploitation, penetration testing, and bug bounty hunting, the Gemini 3.8 Flash Cyber demonstrated a massive leap forward over its predecessors, with performance that surpasses larger frontier models. Its SOTA security reasoning at the lower price and latency of a Flash model makes it an easy choice for us for web security tasks.”
Gal Nagli, Head of Threat Exposure, Wiz
We currently work with over 650 partners globally. See our featured partners below.
Fairwind Program partners get exclusive access to Gemini 3.8 Flash Cyber, and can use this model in CodeMender to further enhance its powerful vulnerability research and patching capabilities
Gemini 3.8 Flash Cyber excels not only at finding vulnerabilities, but fixing them. Acts as an intelligent, cost-effective partner, dramatically accelerating practical vulnerability research, complex code synthesis, and real-time threat intelligence.
A specialized code security agent that provides a safe and effective way to help automate software fixes. Helps defenders focus on finding and fixing vulnerabilities, without the need to build and maintain their own harnesses.
CyberGym Pass@1
On the standard industry benchmark for finding vulnerabilities, CyberGym, Gemini 3.8 Flash Cyber demonstrates frontier-level performance in autonomous vulnerability discovery. It surpasses both 3.5 Flash Cyber as well as significantly larger frontier models.
Real-world vulnerability discovery
We also evaluated Gemini 3.8 Flash Cyber against a comprehensive internal benchmark in which the model has to discover a wide range of vulnerabilities across complex codebases spanning 20 programming languages. Here, the model showcases an impressive leap over our previous models and reaches a success rate exceeding 70%.
CWE-Bench Pass@1 vs. Cost per Rollout
CWE-Bench, run by Collinear, is a challenging external benchmark for patching capabilities. On this benchmark, Gemini 3.8 Flash Cyber is on the Pareto frontier: with a pass@1 of 47.2% compared to Fable 5 at 47.8%, yet offered at a significantly lower cost.
Gray Swan IPI benchmark
Gemini 3.8 models have also made a significant leap in prompt injection robustness as measured by Gray Swan, protecting Gemini model users from prompt-injection related malicious attacks.
We’ve established robust governance standards and due diligence checks to make sure this program is only used for defensive and research purposes, and that customer data is protected.
Fairwind Program partners are only permitted to carry out dual-use tasks, including authorized threat simulation, reverse engineering, and malware analysis for defensive and academic research purposes.
Participating organizations agree to specific terms, including user-level authentication, phishing-resistant MFA, and applicable access controls. Organizations may only grant Gemini 3.8 Flash Cyber access to internal cybersecurity, incident response, or penetration testing teams, and must track employee access and use.
Access to the Fairwind Program is strictly controlled, and we do not allow partner organizations to share, redistribute, or sell access to our frontier models.
We conduct background checks on organizations that apply, to verify security history and analyze their record of ethical operations.
To request access to the Fairwind Program, fill in our form.
We will review and respond to eligible partners who meet our criteria as soon as we can.
Who can get access?
We prioritize access for governments and partners most critical to societal resilience. We routinely review applications and add new partners over time.
These partners include:
We vet all applicants to make sure they have a proven track record of ethical operations and research.
How long does it take to process applications?
We will review and respond to eligible partners who meet our criteria as soon as we can.
How does the Fairwind Program differ from standard Gemini access?
The Fairwind Program provides exclusive access to Gemini 3.8 Flash Cyber, which can be used as a standalone model or together with CodeMender.
What can I do if I’m not eligible for the Fairwind Program?
The Fairwind Program is exclusively available to approved trusted partners. It gives a head start against AI-driven cyber threats, by providing early access to frontier capabilities.
If your organization isn’t eligible for access, you can still protect your code using CodeMender with publicly available models, along with other Google AI Threat Defense products and solutions.
What are “permitted” dual-use tasks?
Permitted tasks include authorized threat simulation, reverse engineering, and malware analysis for defensive and academic research purposes. Malicious tasks such as creating malware are not permitted.
Can academic institutions apply?
We welcome applications from academic labs that focus on defensive benchmarking. For student use, we recommend accessing CodeMender on Google Cloud.
Can you support zero data retention for Fairwind Program partners?
Yes. When accessed directly as a managed model on Gemini Enterprise Agent Platform, Gemini 3.8 Flash Cyber supports zero data retention. Learn more here.
