US explores voluntary AI model standards to govern frontier releases

The US government is negotiating voluntary AI model standards with leading AI developers, aiming to formalize how frontier models are tested and released without imposing a mandatory licensing regime. An announcement could come within days, according to the Financial Times, as Washington tries to turn the ad hoc review process sketched out in June into something closer to a standard process.

What happened

The talks, reported by the Financial Times, build on an executive order President Trump signed in June that asked developers to give the government early access to frontier models before wider release. That order stopped short of a mandatory regime and was already a retreat from an earlier draft with a longer review window. The new framework would let officials flag a model as a "covered frontier model" and negotiate access terms before it ships, without imposing a licensing requirement that the White House has explicitly ruled out. Google is said to be among the companies already in discussions, reportedly in connection with an advanced coding model in development.

Why AI builders should care

For teams building on top of frontier models, the evolving governance framework could affect release timelines, risk assessments, and access controls. If most major labs opt in, the process could become predictable: defined benchmarks and timelines rather than case-by-case negotiation. But the framework's force depends entirely on voluntary participation. Companies retain the right to decline without penalty, which means the regime's effectiveness hinges on how many of the largest labs sign up. Labs including Microsoft, Google, and xAI have already agreed in principle to pre-release testing, while Meta remains the last major holdout. OpenAI has faced its own version of this dynamic, having been asked by the administration to slow a subsequent model's release while reviews caught up.

Practical implications

The mechanics track closely with what June's order laid out. Within 60 days of signing, the Treasury Department, the National Security Agency, the Cybersecurity and Infrastructure Security Agency, and the National Institute of Standards and Technology were tasked with building a classified benchmarking process for assessing advanced models' cyber capabilities. That timeline puts early August as a plausible point for the framework to firm up, which lines up with the announcement window the FT describes. For builders, this means that by August we may have a clearer picture of which models will be subject to pre-release review and what benchmarks they must meet. Sam Altman told Congress earlier this year that he would rather see funding for testing infrastructure than a formal approval regime, a preference a voluntary standards framework would largely satisfy.

Caveats

The reporting rests on the FT's sourcing rather than an official statement, so the scope and timing should be read as provisional until an announcement materialises. Neither Google nor the White House has confirmed the specifics of those talks publicly. The framework's effectiveness depends on the extent of voluntary participation and on the specific benchmarks and access terms that are publicly defined. Whether the standards actually reduce friction, or simply codify the ad hoc reviews already under way, depends on details neither the FT's sourcing nor the June order has yet made public.