Gemini 4 Argon Restricted Over Cybersecurity Misuse Risks

Google is restricting access to Gemini 4 Argon — a new AI model able to autonomously find, validate, and patch software vulnerabilities — to vetted cybersecurity experts while it strengthens safeguards. Early tests (including with partner Wiz) reportedly found a critical vulnerability in healthcare software, Argon scored highly on vulnerability-remediation benchmarks, and Google is addressing risks such as prompt-injection attacks and potential misuse before broader release; pricing and phased availability were also announced.