Overview
- Late June OpenAI unveiled the GPT‑5.6 family and, at the request of U.S. authorities, restricted initial preview access to a small set of trusted U.S. partners.
- The GPT‑5.6 line includes three variants called Sol, Terra and Luna, and OpenAI says Sol can autonomously perform tasks in programming, biology and cybersecurity.
- Before OpenAI’s move the U.S. had ordered Anthropic to limit distribution of its Mythos/Fable models and then approved a partial re-release of Mythos 5 to vetted firms after the company cooperated with regulators.
- A June audit published by journalist Ed Zitron showing a jump in losses has heightened investor concern and several reports say OpenAI is weighing a delay of a planned late‑2026 IPO into 2027.
- Analysts warn that high compute costs and revenue concentrated with a few partners raise sustainability questions, and tighter government controls could steer customers toward open‑source or regional alternatives and reshape global access to frontier models.