The White House Office of the National Cyber Director (ONCD) called a closed-door meeting with OpenAI and other frontier AI labs on May 19, 2026, to brief them on an upcoming executive order on artificial intelligence. The order, expected to be signed by President Trump as soon as Thursday, has been softened: instead of the feared mandatory licensing regime similar to the FDA, it will be a voluntary pre-release review framework. That's a big win for the industry, which had been panicking over the prospect of needing government approval to ship new models.
The new rules come with a timeline twist. Labs will be required to share their frontier AI models with the government up to 90 days before public release — but they're already lobbying hard to cut that to just 14 days. Yes, 14 days. The shorter the review window, the less time the government has to poke around inside their black boxes.
So which models will be subject to review? That definition is being drafted by the NSA, ONCD, and the Office of Science and Technology Policy (OSTP), with security and intelligence agencies joining forces with the Department of Defense to create a classified process for evaluating models that meet the threshold. The immediate trigger for this whole regulatory sprint: Anthropic's decision to withhold its Mythos model after it demonstrated the ability to autonomously discover zero-day vulnerabilities. That was enough to spook Washington into action.