Pull to refresh
Logo
Daily Brief
Following
Why Ranks Sign Up
Trump signs voluntary federal review of frontier AI models

Trump signs voluntary federal review of frontier AI models

Rule Changes

The US government has pulled Anthropic's newest models offline and staged OpenAI's GPT-5.6 launch — going further than Trump's voluntary framework in its first six weeks

August 1st, 2026: Frontier-model benchmark due

Overview

Trump signed a voluntary framework on June 2 asking frontier labs to share their newest models with federal agencies for up to 30 days before release. Within ten days, the government had gone further. A June 12 export control directive forced Anthropic to kill worldwide access to Fable 5 and Mythos 5, its two newest models, after the government said it had found a jailbreak in the more capable one.

On June 26, OpenAI staged the launch of GPT-5.6 at government request, limiting early access to about 20 vetted companies while federal offices reviewed its cybersecurity capabilities. OpenAI said the arrangement should not become the default. For Anthropic's models, the government used an export control directive — no invitation required, no opt-out available.

Why it matters

The government has shown it can halt any frontier model launch — through the voluntary order, export controls, or both.

Questions about this story

No questions yet — be the first to ask.

Key Indicators

30 days
Maximum pre-release review window
Time the federal government may study a frontier model before public release under the voluntary framework.
2 suspended
Frontier models pulled offline by government order
Anthropic's Fable 5 and Mythos 5 were suspended worldwide on June 12 under a US export control directive — the first time the government forced a lab to withdraw deployed models.
Cut from 90
Window in earlier draft
The version the White House planned to sign May 21 gave the government 90 days. Industry pushback shortened it.
3
Labs named as frontier developers
Anthropic, OpenAI and Google are the firms the order treats as primary participants.
60 days
Deadline to define a frontier model
Treasury, NSA, CISA, NIST and White House officials must finish a classified benchmark before any formal review begins. Due August 1.

Voices

Curated perspectives — historical figures and your fellow readers.

Ever wondered what historical figures would say about today's headlines?

Sign up to generate historical perspectives on this story.

Play

Exploring all sides of a story is often best achieved with Play.

Log in to play. Track your picks, climb the leaderboards. Log in Sign Up
Predict 6 ways this could play out. Contrarian picks score more — points lock when the scenario resolves. Log in to play
Higher or Lower Two numbers from this story. Guess which is bigger. 5 rounds to set a streak. Log in to play
Timeline Five events from this story — drag them oldest to newest. Log in to play
Connections Sixteen names from the news. Find the four hidden groups of four. Log in to play

People Involved

Organizations Involved

Timeline

October 2023 August 2026

11 events Latest: August 1st, 2026 Showing 8 of 11
Tap a bar to jump to that date
  1. Frontier-model benchmark due

    Latest Implementation deadline

    Treasury, NSA, CISA, NIST and White House officials must finish the classified benchmark that defines a 'covered frontier model.'

  2. AI cybersecurity clearinghouse due

    Implementation deadline

    Treasury must form a clearinghouse to coordinate AI software vulnerability scanning with industry and critical-infrastructure operators.

  3. US government forces Anthropic to pull Fable 5 and Mythos 5 worldwide

    Policy

    The Commerce Department issued an export control directive requiring Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national, citing a potential jailbreak. Because Anthropic could not segregate access, it suspended both models for all users worldwide. Anthropic disputed the rationale but complied. The models remained offline as of June 26.

  4. Anthropic calls for globally coordinated AI slowdown

    Policy

    Anthropic published 'When AI builds itself,' arguing that AI is accelerating its own development faster than governance can follow and calling for a coordinated option to pause or slow frontier development. David Sacks accused Anthropic of regulatory capture, saying the proposal was designed to protect its proprietary models from open-source rivals. The report came three days after Anthropic confidentially filed for an IPO.

  5. Trump signs scaled-back AI review order

    Policy

    Order asks frontier labs to share models with the federal government for up to 30 days pre-release on a voluntary basis. NSA gets 60 days to set a classified threshold.

  6. Trump postpones AI signing ceremony

    Policy

    Hours before a planned ceremony with tech CEOs, Trump scraps the signing after calls from Musk, Zuckerberg and Sacks objecting to the 90-day review window.

  7. Sacks exits White House AI czar role

    Personnel

    David Sacks steps down as AI czar and moves to PCAST, where he keeps shaping frontier-model policy from outside the building.

  8. AI Safety Institute renamed

    Institutional

    Commerce rebrands the institute as the Center for AI Standards and Innovation, shifting its stated mission from safety to foreign-threat review.

  9. Trump rescinds Biden's AI order

    Policy

    Hours after inauguration, Trump signs EO 14179 'Removing Barriers to American Leadership in Artificial Intelligence,' killing Biden's mandatory reporting regime.

  10. US AI Safety Institute established

    Institutional

    Commerce sets up the AI Safety Institute inside NIST to evaluate frontier models and coordinate with industry.

  11. Biden signs mandatory frontier-AI reporting order

    Policy

    Executive Order 14110 requires top labs to share safety-test results and dual-use research with the federal government.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

October 2023 – January 2025

Biden's frontier-AI reporting order (2023)

On October 30, 2023, President Biden signed Executive Order 14110, the first federal rule requiring top AI labs to share safety-test results and dual-use research with the government. It used the Defense Production Act to make reporting mandatory above set compute thresholds.

Then

Major labs began submitting reports and signing voluntary cooperation agreements with the new AI Safety Institute.

Now

Trump rescinded the order on his first day back in office in January 2025, ending mandatory reporting and leaving a 17-month gap before the June 2026 voluntary framework.

Why this matters now

The 2023 order set the policy reference point the new one is measured against. Biden made disclosure mandatory; Trump has made it optional. The same labs, same models, same risks, two different legal regimes.

June 2025

AI Safety Institute renamed to CAISI (2025)

Commerce Secretary Howard Lutnick renamed the AI Safety Institute as the Center for AI Standards and Innovation. The center's mandate was rewritten to focus on foreign AI threats, backdoors in adversary models and resistance to 'regulatory overreach.'

Then

Safety-focused staff left or were reassigned. Industry kept its existing cooperation agreements but with a less independent counterpart.

Now

By June 2026 CAISI was the technical body the new order had to lean on, but its independent evaluation capacity was thinner than it was under Biden.

Why this matters now

The CAISI rebrand showed which way the administration was pulling on frontier-AI oversight before the new order existed. Safety advocates argue it weakens the body now charged with executing the voluntary review.

2019 – 2024

CFIUS review of TikTok (2019-2024)

The Committee on Foreign Investment in the United States ran a multi-year national security review of TikTok's US operations, ultimately producing a forced-divestiture bill that passed Congress in April 2024.

Then

TikTok continued operating under negotiated security commitments while the review dragged on.

Now

Voluntary cooperation eventually gave way to a statutory mandate when Congress decided the executive branch tools were not enough.

Why this matters now

CFIUS shows how a voluntary or executive-only national security review of a tech product can run for years before Congress writes it into law. The same dynamic could play out with frontier AI.

Sources

(21)