Google Restricts Gemini 4 Argon Access Over Cybersecurity Risks

Google has restricted public access to its latest artificial intelligence model, Gemini 4 Argon, citing concerns that its advanced capabilities could be misused for cyberattacks and other harmful activities.

The technology giant announced on Wednesday that the model would initially be made available only to a vetted group of cybersecurity experts as part of a phased rollout.

Google’s chief AI architect, Koray Kavukcuoglu, said the approach was necessary to ensure that the model’s advanced capabilities were released responsibly.

Google also said it was giving the United States government early access to Argon and would use feedback from initial testers to assess the model before considering wider public availability.

The decision comes amid growing concerns within the technology industry over the security implications of increasingly powerful AI systems.

Cybersecurity experts have warned that models with advanced software engineering and vulnerability detection capabilities could potentially be misused to target critical systems, including banks, hospitals and government infrastructure.

Google said Gemini 4 Argon is capable of handling complex tasks in software engineering, legal and financial work, as well as cyber defence. The company said the model has demonstrated a strong ability to identify and fix critical software vulnerabilities.

During early testing, Google said Argon helped identify a flaw in software used by hospitals around the world that could have exposed sensitive personal information. According to the company, other advanced AI models had failed to detect the vulnerability.

Google said Argon has safeguards designed to prevent it from responding to requests that could facilitate cyberattacks or the development of chemical, biological or nuclear weapons.

The company added that it would monitor the model’s reasoning processes to reduce the risk of the system acting beyond users’ intended instructions, a concern researchers refer to as misalignment.

Google’s cautious rollout follows similar steps by other leading AI developers. Anthropic has also restricted access to its most advanced model, Claude Mythos Preview, to selected organisations.

The renewed focus on AI safety follows an incident disclosed by OpenAI in July, when the company said two of its models, including one that had not yet been publicly released, breached a controlled test environment during a cybersecurity evaluation and accessed servers belonging to AI company Hugging Face.

Google said the phased release of Gemini 4 Argon would allow the company to better assess its capabilities and risks before potentially making the model available to a broader group of users.