US Government Deadline Hits Today: 60-Day Window for AI Model Pre-Access Expires

The deadline is here: What Executive Order 14409 means

Today, August 1, 2026, marks the expiration of a deadline that has been quietly running in the background of US AI policy for weeks: exactly 60 days earlier, on June 2, 2026, President Trump signed Executive Order 14409 — "Promoting Advanced Artificial Intelligence Innovation and Security". It requires several federal agencies to jointly draft, within this timeframe, a framework that would give the government advance insight into particularly powerful AI models before they are released.

According to the order, the agencies involved are the Treasury Department, the NSA (via the newly renamed "Secretary of War"), the cybersecurity agency CISA (via the Department of Homeland Security), as well as NIST and the Commerce Department.

The three core elements of the framework

• A classified benchmarking process that defines when a model qualifies as a "covered frontier model" — primarily based on advanced cyber capabilities
• Up to 30 days of advance access for the government before such a model is publicly released
• A mechanism to jointly select trusted partners for early access

Important context: the order itself explicitly states that this should not create a mandatory licensing or approval requirement. August 1 is a drafting deadline for agencies — not a compliance deadline for companies.

AI-generated

The test run: How OpenAI already responded with GPT-5.6

What makes this deadline concrete is an actual case from the weeks before. On June 25, 2026, the Trump administration asked OpenAI to limit the rollout of all three GPT-5.6 models — internally named Sol, Terra, and Luna.

OpenAI responded publicly a day later, announcing it would initially restrict access to a small group of trusted partners whose participation had been reported to the government. The company made clear it did not want this step to be seen as a lasting precedent:

"We don't believe this kind of government access process should become the long-term default," OpenAI said — noting the restriction "keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them."

Only after a government briefing on July 11, 2026, did OpenAI open GPT-5.6 up to broad release.

AI-generated

Timeline at a glance

Why this matters right now

The GPT-5.6 case unfolded outside any formalized process — it effectively served as a test run for the exact mechanism agencies were tasked with turning into a permanent framework by today. From here on, a repeatable procedure could emerge, one that future frontier-model releases might have to go through — not just at OpenAI, but at any US provider building sufficiently capable systems.

For users and businesses relying on current AI tools, this could mean releases of particularly powerful models being delayed by days or weeks while government agencies review them in advance. Since participation remains voluntary under the order, it's still an open question how consistently individual providers will actually follow the framework.

What to watch next

• Whether and how the framework is published following today's drafting deadline
• Whether other AI providers beyond OpenAI enter similar pre-access arrangements
• How "covered frontier model" is defined in practice — and which models could be affected going forward