Apple isn’t mentioned in this article. I can’t write a headline for Guide4Mac.com because this story is about OpenAI’s GPT-5.6 models and government AI regulation—not an Apple product, service, OS, or feature. Please provide an article about Apple to headline.

OpenAI released three new models under the GPT-5.6 label, but almost nobody can use them yet. The lineup includes Sol (the flagship), Terra (a balanced everyday model with performance similar to GPT-5.5 but at half the cost), and Luna (a fast, low-price option). Access is currently limited to a small group of trusted partners through the API and Codex.

The restricted rollout is not entirely OpenAI’s choice. The Trump administration asked the company to hold back on a broader release, and OpenAI agreed, framing the delay as a short-term concession while the administration works through a cyber executive order framework and develops a repeatable process for reviewing future models. A June 2 executive order formalized the government’s intention to benchmark and assess new AI before it reaches the public.

OpenAI used its own announcement to push back on that arrangement, stating plainly that government access review “should not become the long-term default” and arguing it keeps capable tools away from developers, enterprises, and security defenders who need them. That kind of public friction with the administration, embedded inside a product launch post, is the more unusual part of this story.

The cybersecurity angle

Sol is specifically positioned around agentic work in coding, biology, and cybersecurity, with a new “max” reasoning effort and an “ultra” mode that uses sub-agents for complex tasks. OpenAI says the model is better at helping users find and fix vulnerabilities than at executing end-to-end attacks, and describes the safety stack as its most hardened to date, tested against real-world attack patterns.

The government review process already has one visible casualty. The administration previously forced Anthropic to pull access to Claude Fable 5 and Mythos 5, even after Anthropic completed a voluntary review and added guardrails based on government feedback. OpenAI appears to be watching that precedent closely, complying for now while signaling it expects a faster path to broad availability in the coming weeks.