GOOGLE has unveiled Gemini 4 Argon, its frontier AI model, with a pilot rollout to a restricted group of trusted cyber defenders via its Fairwind Programme, and is gradually expanding access as feedback grows. Argon is pitched as capable of handling complex software engineering, enterprise knowledge tasks, and cybersecurity defence, with Google claiming it can autonomously find, validate and patch critical software vulnerabilities.
For trusted defenders and Google’s internal teams, Argon will be released without cyber guardrails to showcase its full frontier capabilities, while a staged approach remains in place for broader access.
In demonstrations linked to its Scan for Good initiative, Argon reportedly identified a critical vulnerability in hospital software affecting sensitive personal information used by hospitals worldwide, a finding Google says previous frontier models had missed. The company did not name the affected software or indicate whether the issue has been fixed.
Google notes that Argon’s performance has shown improvements in vulnerability discovery versus Gemini 3.8 Flash Cyber and performed strongly in internal benchmarks (CWE-bench v1, where Argon tied first with 68% alongside GPT-6 Astra and Grok 4.7) and Wiz’s black-box testing. The rollout is being carefully controlled, with safeguards such as misalignment mitigations, monitoring of internal activations, and sandbox isolation to prevent misuse as it moves toward a wider release.