US Negotiates Voluntary Standards for Advanced AI Models

Reuters, citing the Financial Times, reports that the US government is in advanced talks with AI companies to create voluntary standards for the release of new models, with an announcement possible as soon as next week. The proposed framework would set benchmarks and timelines for advanced models and clarify who can access them inside the United States and abroad. The FT reporting, relayed by The Next Web, describes a mechanism under which officials could designate a model as a "covered frontier model" and negotiate access terms before release, without imposing a formal licensing regime. Reuters also reports that Google has been part of discussions linked to an advanced coding model, that the Commerce Department recently lifted export controls on Anthropic's Fable and Mythos models, and that OpenAI curtailed a full public launch of GPT-5.6 at the government's request. Reuters said it could not immediately verify the FT report and that several parties did not respond to requests for comment.

1. June 2, 2026

   President Trump signs the "Promoting Advanced Artificial Intelligence Innovation and Security" executive order, directing agencies to work with developers on a voluntary pre-release engagement framework for frontier models.

2. July 1, 2026

   The Financial Times reports the administration is in advanced talks with AI companies on the specific standards, with Reuters saying an announcement could come within days.

3. August 1, 2026

   The voluntary pre-release framework called for in the executive order is due to be finalized, per the order's own implementation timeline.

Voluntary standards of this kind typically hinge on documented evaluation metrics, red-team results, threat models, and staged access plans. For practitioners, that generally means producing standardized risk assessments, reproducible benchmark runs, and clearer access-control matrices ahead of release. The executive order's framework reportedly allows for up to a 30-day early-access window for government review before a model reaches other trusted partners, a concrete window release managers should plan around once the framework is finalized.

• •Whether the government publishes a formal metrics template or submission process ahead of the August 1 deadline.
• •Which agencies issue implementation guidance first, and whether it names specific benchmark suites.
• •Whether companies disclose standardized test artifacts or phased access lists once the framework takes effect.

This is one of the more concrete near-term compliance changes practitioners at frontier labs will face this year, precisely because it now has a legal deadline attached rather than being open-ended policy guidance. Teams shipping models that could plausibly be designated "covered frontier models" should treat the August 1 date as a planning anchor, even though the specific benchmarks and access rules are still being negotiated.