Breaking

US Negotiates Voluntary AI Model Release Standards With Top Labs

July 2, 2026 at 09:01 EDT

Policy & Regulation
Foundation Models
Security

The US government is negotiating a set of "voluntary release standards" for advanced AI models with leading companies including OpenAI, Anthropic and Google. As of July 2, 2026, reports indicate the talks are in an advanced stage, with standards possibly to be announced as soon as next week.

June 2026 · White House Framework

US Weighs Voluntary Standards for Releasing Frontier AI Models

The government is reportedly negotiating a voluntary release framework with OpenAI, Google and Anthropic — offering up to 30 days of early access to "covered frontier models," with an announcement possible as soon as next week.

days of early access offered to the government (maximum)

days to build the classified benchmarking process — by Aug 1

leading firms in talks: OpenAI, Google, Anthropic

Two approaches to frontier-AI oversight

Voluntary

US framework — no mandatory licensing

Mandatory

EU AI Act — binding legal regime

Takeaway: the US emphasizes industry-led, "light-touch" standards — not a legally binding regime.

How a model would be handled

Designate
Classified benchmark of cyber capabilities → "covered frontier model"

→

Early access
Up to 30 days for government & trusted partners

→

Safeguards
NDAs, IP protection, confidentiality, insider-risk measures

The case for it

National-security and cyber-threat defense against misuse of highly capable models
Builds on existing NIST AI RMF and CAISI evaluation tools

The concern

"Voluntary" could become a de facto licensing regime
Uncertain access cutoffs may push firms toward open models
Risk of stifling innovation for IPO-linked OpenAI and Anthropic

Specific benchmarks, access rules and final standards remain under negotiation and unpublished — an official announcement is still awaited.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.