Due to Security Concerns, Google Restricts Gemini 4 Argon

Google has announced the launch of its new artificial intelligence model, "Gemini 4 Argon," under a cautious mechanism, voluntarily granting the US government early access. Koray Kavukcuoglu, the company's Chief AI Engineer, explained that making such leading capabilities available requires adopting a gradual approach to ensure safety, noting that the company will gather feedback from testers before rolling out the model on a wide scale.

Cybersecurity Breach Risks

This caution comes amid growing concerns among cybersecurity experts about the potential exploitation of this advanced technology to breach banks, hospitals, and government systems. Conversely, Google confirms that the "Argon" model excels in completing complex tasks within software engineering, legal, financial, and cyber defense fields, while also possessing superior ability to detect and fix critical software vulnerabilities.

Coordination with the US Administration

This announcement precedes by just one day US President Donald Trump hosting senior executives from tech companies at the White House, where they signed a voluntary agreement committing to monitor risks associated with their companies' AI systems. Google's gradual approach aligns with the strategy of competitor Anthropic, which kept its most advanced model, "Claude Mythos Preview," available to a limited number of trusted organizations.

It is worth noting that Washington had briefly forced Anthropic last June to suspend public access to the "Claude Mythos" and "Claude Veblen" models, before later adopting a voluntary mechanism for auditing the strongest AI models prior to release.