Google keeps Gemini 4 Argon behind a cyber-defense gate

Google is starting Gemini 4 Argon, the AI model it calls its most advanced, with cybersecurity partners rather than the public. The choice reflects a dual-use problem: Google says the system can find, verify, and help fix serious software flaws—useful for defenders, but potentially valuable to attackers too. For now, trusted defenders in its Fairwind program will test it while the company gathers feedback and strengthens protections against misuse before expanding access to developers, companies, and consumers. The rollout shows how more capable AI is changing product launches: the same skills that can secure software can also be why a model is initially kept out of general hands.