What happened
On Sept 30, 2026 Google announced Gemini 4 Argon, its new frontier model, rolling out first to vetted cyber defenders so they can use it 'without cyber guardrails' for defensive work. It can autonomously find, validate and patch vulnerabilities, ties for first on CWE-bench v1 (68%), and Google claims leading robustness on Gray Swan's Indirect Prompt Injection benchmark; it also expands output to a 1M-token limit.
Why it matters
This is the first frontier-class model explicitly trained for and gated to cyber defenders before general availability, signaling AI-for-defense becoming a first-class model capability — Wiz already reportedly used an early instance to uncover a critical vulnerability in global hospital software. It gives SOC/vuln-management teams a frontier-tier defensive model before competitors' equivalent releases and before public access.
Applicability
Enterprise cyber-defense teams and MSSPs in Google's Fairwind program and Google SecOps customers should evaluate Argon for autonomous vuln discovery/patching and secure SDLC workflows now; broader teams should watch for API/enterprise rollout.