US government moves toward case-by-case approval for new AI models

In short: The U.S. government is increasingly controlling when and how some new AI models can be released, including OpenAI’s reported GPT 5.6 rollout.

What happened

A TechCrunch report says the U.S. government is starting to take a much bigger role in deciding which advanced AI models can be released to the public.

Two weeks earlier, the government moved to block Anthropic’s “Fable” and “Mythos” models from a normal public release. Now, according to reporting first published by The Information and cited by TechCrunch, OpenAI’s next model, called GPT 5.6, may only be available in a limited preview. Under that plan, the government would approve access “customer by customer” before allowing a wider release.

OpenAI CEO Sam Altman reportedly suggested the preview could last a couple of weeks. But TechCrunch notes that Anthropic’s Mythos has already been stuck in preview for months, with no clear timeline for a full release.

The article argues that if reviews drag on, it could reduce the financial return on expensive new AI systems. It could also slow down related spending, like building more data centers (large buildings filled with computers that run AI, like factories for computing power).

Why it matters

For regular people, this could affect how quickly new AI features show up in everyday products, and who gets access first. It also raises a bigger question: what risks the government is trying to prevent, and whether it has enough experts to test these systems in a consistent way. The report suggests that companies may need to work together on clearer rules, rather than treating safety reviews as a competitive tool.

Source: TechCrunch AI