GPT-5.6 Preview System Card (deploymentsafety.openai.com)

OpenAI has announced the upcoming launch of GPT-5.6, a new family of three models—Sol, Terra, and Luna—designed to enhance cybersecurity capabilities while emphasizing safety and accessibility. The models will undergo a limited preview with select trusted partners before broader availability in the coming weeks. Notably, the company has developed robust safeguards specifically tailored for each model’s capabilities, particularly in the areas of cybersecurity and biological risks, though none reach the highest risk levels for AI self-improvement. The advancements in GPT-5.6 are significant, as they reportedly improve upon the ability to detect and fix cybersecurity vulnerabilities without enabling harmful autonomous attacks. With new safety features, including activation classifiers and real-time monitoring for unsafe content generation, OpenAI aims to minimize risks effectively. The iterative design process highlights the commitment to safety, with thorough testing to identify vulnerabilities and mechanisms to mitigate potential threats. This strategic focus on improving defensive cybersecurity measures while cautiously regulating offensive capabilities is set to reshape the landscape for AI applications in security and safety.