The Trump administration is in advanced talks with leading AI developers on a voluntary framework governing how frontier models get released, according to a Financial Times report summarized by Reuters, with an announcement possible as soon as next week. The framework would set benchmarks for advanced models, define release timelines, and clarify who can access such systems inside the United States and abroad.

The context is a quiet tightening. Washington has grown more anxious about frontier systems reaching military intelligence services in China, Russia, or other countries of concern, and it has been converting that anxiety into structure rather than statute.

That structure took formal shape on June 2, when President Trump signed an executive order titled “Promoting Advanced Artificial Intelligence Innovation and Security.” The order directs agencies to build a classified benchmarking process within 60 days and establishes a voluntary window of up to 30 days before release during which developers can give the federal government access to covered frontier models. The text also explicitly states that nothing in it authorizes a mandatory licensing, preclearance or permitting regime. The design is unmistakable: the leverage of preclearance without its legal footprint.

The consequences are already visible. On June 26, OpenAI announced it was limiting the preview of GPT-5.6 Sol, Terra and Luna to a “small group of trusted partners whose participation has been shared with the government,” describing the restriction as a “short-term step” on the path to broader availability in the coming weeks. Earlier, the administration ordered Anthropic to strip foreign-national access from its Fable 5 model; Anthropic took the model down entirely, per TechCrunch.

The White House, OpenAI and Anthropic didn’t respond to Reuters’ requests for comment. The silence tracks. A voluntary regime works precisely because nobody wants to describe it out loud.

Sources