How to Apply for Fairwind Program Access to Gemini 4 Carbon
Step-by-step guide to apply for Google’s Fairwind Program and request early access to Gemini 4 Carbon for cybersecurity defense work.
Google released Gemini 4 Carbon on 28 October 2026 as its strongest frontier model, with leading results on cybersecurity benchmarks. Access starts with a subset of cyber defenders this week and expands through the Fairwind Program before broader release.
If your organization defends critical systems, the Fairwind Program is the current route to early Carbon access. This guide covers eligibility, the application process, requirements after approval, and what to do while you wait.
What is the Fairwind Program?
The Fairwind Program gives trusted partners early access to Google’s most capable cyber defense models. It prioritizes organizations that protect societal resilience, such as governments and operators of essential services.
Google currently works with more than 650 partners worldwide. Partners receive models with cybersecurity capabilities, including the option to use them with CodeMender, Google’s agent for finding and fixing vulnerabilities.
Carbon follows the same pattern as Gemini 4 Argon. Argon reached Fairwind partners first, then became available more widely. Carbon is following the same deliberate rollout while Google gathers feedback on guardrails and participates in U.S. government voluntary pre-release testing.
Who can apply?
Google prioritizes these groups:
- Governments and national cyber authorities that defend public-sector networks and citizen services.
- Critical infrastructure operators in healthcare, telecommunications, energy, and financial networks.
- Core technology platforms whose software underpins large numbers of users.
- Academic labs focused on defensive benchmarking.
Google conducts background checks on applying organizations to review security history and ethical operations. Student use is directed to CodeMender on Google Cloud rather than full Fairwind access.
If your organization does not fit these categories, you can still use CodeMender with publicly available models and other Google AI Threat Defense tools.
How to apply for Fairwind Program access
The application is straightforward but selective.
- Go to the official Fairwind Program page at deepmind.google/fairwind-program/.
- Click the Apply now button or the form link.
- Fill in the interest form with details about your organization, its role in cybersecurity or critical infrastructure, and how you plan to use the models for defensive work.
- Submit the form. Google reviews applications on a rolling basis and responds to eligible partners as soon as possible. There is no published timeline.
After submission, prepare documentation that demonstrates your organization’s security practices, incident response capabilities, and ethical track record. These help support the background check.
Existing Google Cloud customers can also contact their account team to discuss eligibility.
Requirements after approval
Approved partners must meet specific security and usage conditions:
- Use phishing-resistant multi-factor authentication.
- Limit model access to internal cybersecurity, incident response, or penetration testing teams.
- Track which employees use the models and for what purposes.
- Restrict use to permitted dual-use tasks: authorized threat simulation, reverse engineering, and malware analysis for defensive or academic research. Creating malware or other malicious activity is prohibited.
- Do not share, redistribute, or sell access.
When accessed as a managed model on Gemini Enterprise Agent Platform, zero data retention is supported.
These rules keep powerful capabilities focused on defense.
What Gemini 4 Carbon offers cyber defenders
Carbon ranks number one on the Artificial Analysis Intelligence Index v5.0 at 59. On cybersecurity benchmarks it leads in finding real vulnerabilities (88.2% on SEC-bench Pro), validating them through exploitation (54.8% on ExploitGym), and producing working patches (71.0% on CWE-bench).
Internal results show more than double the success rate of Argon on generating secure code. Google’s Chrome team reported substantially better high-quality patches compared with another large frontier model. The model is designed for long-horizon tasks that can run for days while tracking down and fixing subtle bugs across large codebases.
Introductory pricing matches Argon at $2 per million input tokens and $10 per million output tokens, with cached input at a 95% discount. After the introductory period, prices rise to $4 and $20.
For context on Argon’s earlier Fairwind rollout, see our guide on applying for Gemini 4 Argon Fairwind access.
Tips for a stronger application
Focus on concrete defensive impact. Describe specific systems you protect, past vulnerability research or incident response work, and how early access to a stronger model would improve outcomes.
Highlight existing security controls such as MFA already in place, access logging, and clear separation between security and other teams.
If you are an academic lab, emphasize defensive benchmarking plans and any prior responsible disclosure work.
While waiting, continue using Argon (available now on the paid API and for Google AI Pro/Ultra subscribers) and CodeMender with public models so your workflows are ready when Carbon arrives.
What happens next
Google is expanding Carbon access gradually through Fairwind. Broader availability for developers, businesses, and consumers is described as coming as soon as possible after more guardrail feedback.
No public API model ID or Gemini app toggle exists for Carbon yet. Check the official Gemini models page and Fairwind Program page for updates.
Organizations that apply now position themselves for the next wave of access. Those that do not qualify can still strengthen defenses with the tools already available to everyone.
Sources
- Google blog: Introducing Gemini 4 Carbon (28 October 2026)
- Google DeepMind: Fairwind Program page
- Google blog: Fairwind Program launch (September 2026)
Share this story