What happened
On June 26, 2026, OpenAI announced a limited preview of GPT-5.6 (comprising three models: Sol, Terra, and Luna) at the request of the Trump administration, releasing only to a small group of government-vetted partners before broader availability. GPT-5.6 Sol was described as a next-generation model with stronger capabilities in coding, science, and cybersecurity, paired with OpenAI's most advanced safety stack. General availability is expected in coming weeks.
Why it matters
The government-gated release establishes a new precedent — frontier AI models with offensive/defensive cybersecurity capability are now subject to national security vetting before public release, paralleling Anthropic's Mythos export controls. CISOs planning to integrate next-generation AI into security operations must account for access unpredictability.
Applicability
Enterprise security teams planning AI roadmaps should model access uncertainty into procurement; vetted Daybreak/government partners will get early access. Broader GA expected within weeks.