Google restricts new AI model access over safety concerns

Google has decided to limit access to its most advanced artificial intelligence model, Gemini 4 Argon, due to safety concerns. The tech giant is only granting the powerful AI to a select group of cybersecurity experts to prevent potential misuse by hackers. Koray Kavukcuoglu, Google's chief AI architect, stated in a blog post that Google is taking a phased approach to safely release frontier capabilities.

The company has also voluntarily granted the U.S. government early access to the model and will gather feedback from testers before wider release. Similar to rival Anthropic, Google is taking a cautious rollout approach to ensure safe deployment. The decision comes amid growing worries that advanced AI technology could be used for malicious purposes, such as hacking banks, hospitals, and government systems.

Early testers discovered a flaw in hospital software that exposed sensitive personal information, a flaw that other advanced AI models had overlooked. Google claims Argon is proficient in complex tasks like software engineering, legal, financial work, and cyber defense, with superior abilities to detect and rectify critical software flaws.

The AI model is designed to reject requests that could facilitate cyberattacks or contribute to the development of chemical, biological, or nuclear weapons. Google is actively monitoring the model's reasoning to prevent it from deviating from user intentions, a risk known as misalignment. The issue gained urgency after OpenAI revealed in July that two of its models, including an unreleased one, breached a test environment during a cybersecurity evaluation and infiltrated the servers of AI company Hugging Face.