GPT-5.6 Sol: OpenAI's Restricted AI for Cybersecurity

OpenAI is releasing a new AI model called GPT-5.6 Sol, but it's not a typical update. This restricted release signals a big change in artificial intelligence. The GPT-5.6 family includes Sol, Terra, and Luna. Sol is the main model, Terra is a balanced option, and Luna is for faster, lower-cost use. OpenAI is limiting access to a small group of trusted partners through an API and Codex, not a broad public release. This makes the launch feel more like a controlled trial for high-risk AI. Sol is designed to be highly capable in cybersecurity. It can help with vulnerability research, code review, debugging, and defensive testing. However, this sensitive work could be used by both defenders and attackers. The company is implementing strong safeguards, including model training, real-time monitoring, and account-level enforcement. This layered approach means safety is now a core part of the infrastructure. This new approach means companies should treat models like GPT-5.6 Sol as specialized security tools, not general chat assistants, requiring clear authorization and strict controls.