Google restricts access to new AI model due to security concerns - Sarmad

Google announced that it is temporarily withholding its Gemini 4 Argon artificial intelligence model from the general public, making it available exclusively to a select group of cybersecurity experts, in a move aimed at preventing misuse by internet hackers.

Koray Kavukcuoglu, Google’s chief AI engineer, wrote in a blog post announcing the model’s release: “Safely providing leading capabilities at this level requires a phased approach.”

According to the Associated Press, Google said it would voluntarily grant the U.S. government early access to the model and would collect feedback from testers before making it widely available.

This cautious rollout aligns with the approach taken by rival Anthropic, which has kept its most advanced model, Claude Opus Preview, available to a limited number of trusted organizations.

In June, Washington briefly forced Anthropic to suspend public access to its Claude Opus and Claude Haiku models, before later adopting a voluntary mechanism for auditing the most powerful AI models before their release.

The announcement comes a day after U.S. President Donald Trump hosted top executives from technology companies at the White House, where they signed a voluntary agreement pledging to monitor the risks associated with their companies’ AI systems.

Cybersecurity experts fear that this advanced technology could be used to hack banks, hospitals, and government systems.

Google stated that the Argon model excels at complex tasks in software engineering, legal and financial services, and cybersecurity, and possesses exceptional ability to detect and remediate critical software vulnerabilities.