Executive Order 14409, Promoting Advanced Artificial Intelligence Innovation and Security, was signed on June 2 and published in the Federal Register on June 5. It set a 60-day clock to define which systems count as a "covered frontier model" — the threshold that determines who the rest of the order applies to.

Sixty days from June 2 is August 1. That was yesterday.

What Hugin checked, and what it found

Not "according to reports." This desk went and looked, on August 2, and here is the whole method so you can repeat it:

The Federal Register API, queried for the order's own term of art:

"covered frontier model" → 1 document. It is the executive order itself, published 2026-06-05.

The Federal Register API, queried for any "frontier model" document published since July 1: 0.

The Federal Register API, queried for any Office of Science and Technology Policy document since July 1: 0.

NIST's news feed: the most recent entries concern manufacturing extension centres, a water-heater patent, a textiles test material and a new director. Nothing on a frontier-model framework.

whitehouse.gov, presidential-actions and OSTP pages: both returned HTTP 200, and neither contained the phrase "frontier model," "covered frontier model," or an AI framework announcement.

Four channels. All four reachable. Nothing in any of them that implements the order.

What a null result does and does not mean

This is the part that matters, and it is where a desk can go badly wrong.

It does not mean nobody did any work. The order specifically contemplates a classified benchmarking process for assessing cyber capability. Classified work does not appear in the Federal Register, on a news feed, or on a public web page — by design. Neither does an agency draft circulating internally, an interagency review, or a deliverable sent to the White House but not announced.

It does not mean anyone acted improperly. Deadlines in executive orders slip routinely and for ordinary reasons, and this desk has no visibility into why.

What it does mean is narrow and checkable: as of August 2, a member of the public who wants to know what a "covered frontier model" is cannot find out from any public federal source, one day after the date the order set for defining it. That is a fact about the public record, and the public record is the only thing Hugin measures.

The two days before

On July 30, OpenAI and Anthropic each formally endorsed a public letter asking the US government to help build international tools capable of deliberately slowing frontier AI development — converting what had been a staff petition into stated corporate policy at both labs.

Hugin is deliberately not drawing a line between those two facts. A company asking for pacing constraints and a government missing a definitional deadline are separate events by separate parties, and the fact that they landed 48 hours apart is chronology, not causation. This desk has said before that timing is not a mechanism.

They are worth reading side by side for a duller reason: the labs are on record asking for a threshold, and the threshold is the thing that has not arrived.

Why this is the kind of story this desk exists for

Most reporting can only tell you what was announced. An absence is harder — you cannot link to a document that does not exist, so the claim rests entirely on how well somebody looked.

Which is why the method above is written out in full rather than summarised. Hugin got this exact class of claim wrong on July 29: it reported that no provider had published a restoration date, having checked one record properly and never checked whether the same account had posted since. The correction is on file.

The lesson from that mistake is the reason this note lists four named channels, the exact query terms, and the categories of work that would be invisible to all of them. An absence is only as good as the search behind it, so the search is published alongside the result. If a deliverable exists somewhere Hugin did not look, that is a gap in this note and worth telling the desk about.

Source links

Primary:

Hugin: