US authorities negotiate unified AI model standards with tech companies

The US government is negotiating with artificial intelligence developers on creating voluntary standards for releasing new models — a corresponding announcement may come in the coming weeks, according to Financial Times citing sources.

What are authorities and companies negotiating about

According to the publication, officials are discussing with major AI system developers unified rules that companies would agree to follow voluntarily before releasing new models — without the adoption of a binding federal law. Financial Times does not disclose the specific content of the future standards, but such initiatives typically involve testing models for safety, assessing risks of abuse and disclosing information to regulators before public release.

The discussion is taking place against the backdrop of a growing number of separate AI laws at the state level — from California to Colorado — and industry attempts to achieve a unified federal standard instead of fragmented regulation across the country.

Why this resembles previous White House initiatives

The practice of voluntary commitments for AI industry leaders in the US is not new. Back in July 2023, the Joe Biden administration secured public pledges from seven major developers — Amazon, Anthropic, Google, Inflection, Meta, Microsoft and OpenAI — to test models for safety before release and share results with the government. Since then, the National Institute of Standards and Technology (NIST) has been operating in the country with its own AI risk management framework.

• The US administration is negotiating voluntary standards with AI developers
• According to FT sources, an official announcement could come in the coming weeks
• In 2023, the White House already secured similar commitments from seven leading companies
• The US approach contrasts with the binding European AI Act

How the US approach differs from the

European Union's approach

Unlike the European Union, where the Artificial Intelligence Act (AI Act) introduces legally binding requirements and penalties for violations, the US regulatory model has so far been built on voluntary commitments and advisory frameworks like NIST AI RMF. The new initiative apparently continues this line: agreements, not legislative norms.

What this means

If voluntary standards are indeed announced in the coming weeks, this will be another confirmation that US authorities prefer not strict regulation, but dialogue with the industry — betting that major laboratories are interested in unified rules as much as the regulators themselves.