Google introduced Gemini 4 Argon, but gave access only to cybersecurity professionals

Google introduced Gemini 4 Argon, but gave access only to cybersecurity professionals
Photo: Gemini app on Google Play / unsplash

Google has officially introduced its new flagship AI model, Gemini 4 Argon. However, at the first stage, the company is not opening it up to a wide range of users and developers – access is granted only to a limited group of vetted cybersecurity experts.

The model is distributed through the Fairwind program, Google said. The company explains the phased launch by the need to further verify the model's protection and reduce the risk of using its capabilities for cyberattacks.

Gemini 4 Argon is designed for complex and long-running tasks that require multi-step analysis. Google highlights three main areas: software engineering, work with corporate information – particularly in the financial and legal spheres – and cybersecurity.

The company pays special attention to cyber defense. According to TechCrunch, Google states that Argon can autonomously find, verify, and help fix critical vulnerabilities in software. Such capabilities at the same time increase the risk of abuse in case of uncontrolled access.

Therefore, Google is initially giving the model to professionals who are engaged in protecting information systems. In parallel, the company is participating in a voluntary U.S. government procedure that involves access to the most powerful models before their wide release. Feedback from early users is expected to help refine limitations and security mechanisms.

Google has not yet named a specific date for the general availability launch. The company only states that it plans to provide Argon to developers, enterprise customers, and general users after the first phase of testing is completed.

Within Google itself, Gemini 4 is already being used for programming and engineering tasks, including finding bugs and migrating large codebases. The company also claims high results for Argon in a number of tests, but these indicators still need to be confirmed through broad practical use.

As Kurs Ukrainy previously wrote, Bloomberg had reported mixed assessments of Gemini 4 inside Google. Some employees believed that high test results do not always translate into the same quality on real tasks, particularly in programming. Google disagreed with that assessment.

Thus, the official launch became a new stage after earlier reports about internal testing of Gemini 4: the model received the name Argon and has already been handed over to external partners, but Google is delaying its wide distribution for now.

Based on materials: Google, TechCrunch