Google is keeping its most powerful artificial intelligence model away from the general public for now. The company said on Wednesday that its new Gemini 4 flagship, codenamed Argon, will be available only to a vetted group of cybersecurity experts and the US government.
The decision marks one of the most cautious releases yet for a frontier AI model, and signals how seriously leading labs now treat the risk that state-of-the-art systems could be turned into hacking tools.
“Safely releasing frontier capabilities at this level requires a phased approach,” wrote Koray Kavukcuoglu, Google’s chief AI architect, in a blog post announcing the model.
A Phased Rollout for a Powerful Model
Google said it is voluntarily giving the US government early access to Argon and will gather feedback from the trusted testers before making the model widely available. The company has not set a timeline for a broader public release.
The announcement came a day after President Donald Trump hosted the heads of Google, Anthropic and OpenAI at the White House, where executives including Google chief Sundar Pichai and Anthropic's Dario Amodei signed a voluntary accord pledging to police the risks of their own AI systems.
Echoes of Anthropic's Restricted Strategy
Google's cautious rollout mirrors the approach of rival Anthropic, which has kept its most advanced model, Claude Mythos Preview, restricted to a small number of trusted organisations rather than releasing it to the public.
Washington has been pushing labs to tighten their release practices. The US government briefly forced Anthropic to suspend access to its publicly released Claude Mythos and Claude Fable models in June, and has since set up a voluntary process for vetting the most powerful AI models before release.
What Argon Can — and Refuses to — Do
According to Google, Argon excels at complex tasks in software engineering, legal and financial work, and cyber defence, with a leading ability to find and fix critical software flaws.
Early testers used the model to uncover a flaw in software used by hospitals around the world that exposed sensitive personal information — something other advanced models had missed, the company said.
Those same capabilities are what make the model dangerous in the wrong hands. Cybersecurity experts fear the technology could be used to hack banks, hospitals and government systems.
Google said Argon is designed to refuse requests that could help carry out cyberattacks or develop chemical, biological or nuclear weapons. Anthropic and OpenAI have built similar safeguards into their most advanced models, though researchers continue to probe the limits of those defences.
As frontier models grow more capable, the Argon release may set a template for how labs balance competitive pressure against safety — with their most powerful systems reaching defenders and regulators first, and the public only later.
