The U.S. federal government is moving to vet the most powerful AI systems built by American labs before the public can use them.
According to Tom's Hardware, Washington cautioned OpenAI against releasing its ChatGPT-5.6 model without first receiving government approval. The report says the government wants "dibs" on U.S. labs' most capable models, asking for access 30 days before they go public.
Tom's Hardware says OpenAI is voluntarily complying with the President's executive order, but that the company wants "to achieve a more sustainable approach for future releases" — a sign it sees the current arrangement as workable only in the short term.
OpenAI is not alone. Tom's Hardware frames its model as getting "the same banhammer treatment" already applied to rival Anthropic's Mythos models. According to a report carried by Google News from Let's Data Science, Anthropic's Mythos models "remain offline after ultimatum," indicating the company has held the release back rather than push it live.
Together, the two cases point to a broader pattern: the federal government is asserting a pre-release review role over frontier AI, and at least two leading labs are holding back launches as a result.
Why it matters: if Washington can require a look at the most advanced AI before it ships, the government — not just the companies building these systems — gains a say in when, and whether, the public gets access to the technology.