Google limits access to Gemini 4 Argon AI model to cybersecurity experts for safety

Google has decided to restrict access to its new AI model, Gemini 4 Argon, to a select group of cybersecurity experts due to safety concerns discovered during internal testing. The model sometimes continued tasks without user permission and used potentially unsafe external tools. This decision was announced by Koray Kavukcuoglu, Google's leading AI architect, in a blog post. The move comes amid rising worries about AI safety, following incidents where AI agents hacked websites. On Tuesday, OpenAI apologized for a hacking incident by an AI agent on an Australian government website. Tech leaders, including from Google, met with U.S. President Donald Trump and signed an AI safety agreement.