Seen in hindsight, the executive order signed on June 2 is the document that organised the rest of the semester in artificial intelligence regulation.
It created the pre-release review structure that, ten days later, would result in two frontier models being pulled from the market, and that would keep a third restricted to approved organisations for weeks before public release.
It also produced the reaction: more than a hundred security leaders signing a petition against the removal, a congressional response deadline expiring with no public statement, and a large institutional buyer testing competitors to replace the unavailable model.
The period's balance shows both sides of the design. On one hand, a mechanism able to act quickly in the face of declared risk. On the other, an absence of predictability and of a clear route of appeal for anyone with a product on the market.
For anyone building on someone else's model, the practical conclusion doesn't depend on which side is right: supplier availability became a regulatory variable, and regulatory variables are managed like any dependency, with a qualified alternative tested in advance.