The White House was in advanced talks with OpenAI, Google and Anthropic over voluntary standards for releasing frontier models, according to reporting from the period. The scope under discussion covers capability benchmarks, testing timelines before release, and access rules.
The word voluntary does heavy work in that sentence. A standard negotiated between government and the three largest companies tends to become a market reference even without force of law, because anyone outside it has to explain why, and because corporate buyers start requiring adherence in contracts.
Who sits at the table matters too. A standard written with the three largest tends to reflect what they can already meet, favouring those with testing infrastructure in place and creating an entry cost for anyone starting out.
The contrast with Europe is direct: there the requirement arrived as law, with obligations to document and publish a summary of training material. Two distinct models of regulation competing to become the international reference.
For anyone building on someone else's model, the practical effect of either route is the same: more available information about what the model is, and more ability to require it in a contract.