US government in talks with AI companies over common model standards

The US government is negotiating with artificial intelligence developers over voluntary standards for releasing new models, and an announcement may come in the next few weeks, the Financial Times reports, citing sources.

What the government and companies are negotiating

According to the FT, officials are discussing a common set of rules with major AI developers. Companies would agree to follow these rules voluntarily before releasing new models, with no binding federal law involved. The Financial Times does not disclose what the future standards will contain, but initiatives like this typically cover safety testing of models, assessment of misuse risks, and disclosure of information to regulators before a public release.

The talks come as a growing number of individual states, from California to Colorado, pass their own AI laws, and as the industry pushes for a single federal standard instead of a patchwork of regulations across the country.

Why this resembles earlier White House initiatives

Voluntary commitments from AI industry leaders are nothing new in the US. Back in July 2023, the Joe Biden administration secured public pledges from seven of the largest developers — Amazon, Anthropic, Google, Inflection, Meta, Microsoft, and OpenAI — to safety-test their models before release and share the results with the government. Since then, the National Institute of Standards and Technology (NIST) has operated its own AI risk management framework.

• The US administration is negotiating voluntary standards with AI developers
• According to FT sources, an official announcement may come in the next few weeks
• In 2023, the White House already secured similar commitments from seven leading companies
• The US approach contrasts with the EU's binding AI law (the AI Act)

How the US approach differs from the EU's

The European Union's AI Act imposes legally binding requirements and fines for violations. The US, by contrast, has so far built its regulatory model on voluntary commitments and advisory frameworks such as the NIST AI RMF. The new initiative appears to continue that approach: agreements rather than legislation.

What it means

If voluntary standards are indeed announced in the coming weeks, it will further confirm that US authorities prefer dialogue with the industry over strict regulation. The bet is that major labs want common rules of the game just as much as regulators do.