NEWS | Google announced Gemini 4 Argon on Wednesday, September 30, a new top-tier model that anchors its Gemini 4 line and is larger than its earlier Pro models. The public cannot use it yet. Access is limited to vetted cybersecurity experts and Google's own teams through a program called Fairwind, and no date has been given for a wider release.
Google says Argon handles software engineering, legal and financial work, and cyber defence. Early testers used it to find a flaw in software used by hospitals worldwide that exposed personal data. The model is built to refuse requests that help with cyberattacks or weapons, and the US government gets early access under a voluntary review process.
The launch came a day after tech leaders signed a White House pledge to police their own AI systems. It also matches Anthropic's choice to keep its most advanced model with a small group of trusted organizations. Washington briefly forced Anthropic to suspend two public models in June, and a voluntary vetting process followed.
BMN view. Everyday users will not notice a change today, since current Gemini tools stay the same. Still, the banks, hospitals and government systems that experts worry about serve everyone. Giving defenders the strongest AI first is becoming a pattern, and Google also faces pressure after months of delays.
If Argon is too risky for the public today, who decides the day it suddenly is not?
