Google launches Gemini 4 Argon for vetted cyber defenders

Google has announced Gemini 4 Argon, a new frontier AI model that is initially being made available to a select group of trusted cyber defenders through its Fairwind Program. Google’s own internal teams are also using the model, and access will be widened gradually as feedback comes in from early testers.

The company says Argon is built for complex workflows in software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defence. On the security side, it can autonomously find, validate and patch critical software vulnerabilities, and Google is releasing it without cyber guardrails for trusted defenders and internal teams so they can use its full frontier-level capabilities.

Google says Argon uncovered a critical vulnerability exposing sensitive personal information in healthcare software used by hospitals worldwide, a risk that previous frontier models had missed. The affected software was not named and Google did not say whether the issue has been fixed. Wiz, which Google acquired earlier this year, is using Argon in its Scan for Good initiative, which finds and remediates high-risk exposures in critical public infrastructure for free.

On CWE-bench v1, a vulnerability remediation benchmark from Collinear AI, Argon tied for first place with a score of 68%, alongside OpenAI’s GPT-6 Astra and xAI’s Grok 4.7. Google also claims gains over Gemini 3.8 Flash Cyber in vulnerability discovery, and says Argon found exposures across complex codebases written in 20 programming languages on its internal benchmark, and outperformed Gemini 3.8 Flash Cyber on Wiz’s black-box penetration testing benchmark.

Ahead of a wider release, Google says the model is designed to refuse requests that could enable cyber or chemical, biological, radiological and nuclear attacks while still supporting legitimate dual-use research. Safeguards include monitoring of the model’s internal activations, misalignment mitigations that watch its chain-of-thought and actions, and isolation of sandboxed environments before high-risk training or evaluations. Google says Argon is its most resilient model yet against indirect prompt injection, and that it is taking part in the US government’s voluntary pre-release model access process.