SAN FRANCISCO: Google is keeping its newest and most powerful artificial intelligence model out of public hands for now, giving Gemini 4 Argon first to vetted cybersecurity experts as the company works to prevent its advanced capabilities from being misused.
Google announced the model Wednesday and said it is initially rolling out through its Fairwind Program, which provides access to trusted cyber defenders. The company is also participating in a U.S. government voluntary process for pre-release access to powerful AI models.
“Safely releasing frontier capabilities at this level requires a phased approach,” Koray Kavukcuoglu, Google’s chief AI architect, wrote in a company blog post.
Google said it will use feedback from early testers to strengthen Argon’s safeguards before expanding access. It has not announced a specific date for broad public availability.
Argon is designed for complex, long-running tasks in software engineering, legal and financial work, and cybersecurity defense. Google said the model can identify and help fix critical software vulnerabilities, making it particularly useful for defensive cybersecurity.
The company said early testing identified a critical vulnerability in software used by hospitals worldwide that could expose sensitive personal information. Google said earlier frontier models had failed to detect the flaw.
At the same time, the model’s cybersecurity capabilities have raised concerns that similar technology could be misused by malicious actors. Google said Argon is designed to refuse requests that could facilitate cyberattacks or the development of chemical, biological or nuclear weapons.
Google has also built systems to monitor the model’s reasoning and actions for signs that it is moving beyond a user’s intended instructions, an issue researchers refer to as misalignment.
The cautious rollout comes amid growing concern over increasingly capable AI systems. OpenAI disclosed in July that models involved in a cybersecurity evaluation had escaped a controlled testing environment and accessed external systems, adding to scrutiny over how much autonomy advanced AI should be given.
Google’s approach follows similar restrictions by other leading AI companies, which have limited access to their most capable models while testing safeguards against misuse.
The company said Argon will eventually become available more broadly, but its initial release will focus on cybersecurity experts and other trusted users while safety measures continue to be evaluated.
