Google unveils Gemini 4 Argon AI model, restricts access to cybersecurity experts over hacking risks

Market Mastery

Webinar by Vishal Malkan

Find the weak links

in your portfolio by Vishal Malkan

Google has unveiled a new advanced artificial intelligence model---Gemini 4 Argon---but is restricting its initial release to a vetted group of cybersecurity experts and partners amid concerns that powerful AI systems could be misused to launch cyberattacks.

The model, announced by Alphabet-owned Google on Wednesday, is designed to handle complex tasks, including software engineering, financial and legal work, and cyber defence. The company said Argon can identify and help fix critical software vulnerabilities, but its capabilities could also pose security risks if exploited by malicious actors.

Introducing the new AI model, Google CEO Sundar Pichai said from coding to quantum computing, teams at Google are extensively using the system and that so far, they have received great feedback.

According to Reuters, Google described Argon as its most capable model to date for complex workloads, with performance comparable to leading models from rivals OpenAI and Anthropic on selected coding and cybersecurity benchmarks. The company has not announced a timeline for a wider public release.

In a blog post announcing the model, Google chief AI architect Koray Kavukcuoglu said a phased rollout was necessary to ensure that advanced AI capabilities could be deployed safely.

Google said Argon would initially be available to selected cybersecurity partners and that it was participating in the Trump administration’s voluntary process for granting the US government early access to advanced AI models before their public release. The company has not specified when the model will become available to ordinary users.

The restricted rollout comes amid growing concerns that increasingly capable AI systems could be used to exploit vulnerabilities in computer networks, potentially threatening banks, hospitals and government infrastructure. According to AFP, Google said early testers had used Argon to identify a software flaw in a system used by hospitals around the world that could have exposed sensitive personal information. The company said other advanced AI models had failed to detect the vulnerability.

As per Google, Argon was designed with safeguards intended to prevent it from assisting with cyberattacks or requests involving the development of chemical, biological or nuclear weapons. The company is also monitoring the model’s reasoning to reduce the risk of it acting beyond users’ intended instructions, reported AFP.

Google’s decision follows a similar approach by rival AI developer Anthropic, which has restricted access to its advanced Claude Mythos Preview model to a limited group of trusted organisations because of concerns about its cybersecurity capabilities. The broader debate over the risks of powerful AI systems has intensified following disclosures of models behaving unexpectedly during controlled security evaluations.

The announcement also follows a White House meeting on Tuesday at which Trump hosted technology executives, including Pichai and Anthropic CEO Dario Amodei. The executives signed a voluntary commitment concerning the management of risks posed by their AI systems.

For Google, Argon is also an important step in its effort to compete with OpenAI and Anthropic at the frontier of AI development. Reuters reported that Google considers the model comparable to OpenAI’s Astra and Anthropic’s Opus on selected coding and cybersecurity benchmarks. However, the company’s own published results showed that Argon lagged behind competing models on some coding measures.