Google Limits Gemini 4 Carbon To Vetted Cyber Defenders

Google announced Gemini 4 Carbon on October 28 and gave it first to a subset of vetted cyber defenders rather than to developers or consumers. The model arrives about two weeks after Gemini 4 Argon. Google says it finds vulnerabilities, writes exploits and patches code better than Argon does, and it did not give a date for broader release.

For security leaders the release matters because a second frontier lab has decided that its strongest cyber model is too risky to ship to everyone. The rules Google sets for who gets in will shape the terms your own security team is offered. Google has not yet published the full list of who is inside.

How Fairwind Controls Access

Carbon reaches defenders through Fairwind, a program Google says works with more than 650 partners. Governments, critical infrastructure operators and core technology platforms get priority, and academic labs focused on defensive benchmarking can apply. Organizations fill out an interest form and Google runs background checks on their security history.

Partners must use phishing-resistant multifactor authentication, limit the model to internal security, incident response and penetration testing teams, and track which employees use it. Reselling or sharing access is not allowed. Permitted work includes threat simulation, reverse engineering and malware analysis, while building malware is prohibited.

The model itself is built for long jobs. Google says Carbon has run for days inside its own teams on tasks such as tracing a bug to its root cause in a large codebase and then patching it. Defenders can aim it at a legacy codebase and let it work through the backlog, with CodeMender, Google’s code security agent, available to automate fixes.

Google reports 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on CWE-bench for patching. It also says the model more than doubles Argon’s success rate at generating secure code.