Frontier AI models should go to trusted cyber defenders before public release

Google's next-generation frontier AI model, unveiled on September 30, 2026 and first released only to trusted cyber defenders through Google's Fairwind Program, ahead of wider availability.

Google DeepMind's frontier AI model, released on October 28, 2026, two weeks after Gemini 4 Argon's wide rollout, first only to selected cybersecurity defenders through Google's Fairwind Program; it ranked first on the Artificial Analysis Intelligence Index v5 with a score of 59.