On June 2, 2026 — seven days before Anthropic launched Fable 5 — President Trump signed an Executive Order titled "Promoting Advanced Artificial Intelligence Innovation and Security."
The order has been largely overshadowed in AI coverage by the Fable 5 ban story. It should not be. Section 3 of this EO is the single most important document for understanding why the ban happened, what the government actually wants, and what the resolution path for Fable 5 actually looks like.
The short version: the EO mandated a classified benchmarking process to designate AI models as "covered frontier models" and a voluntary framework giving the government 30 days of pre-release access before developers ship those models to other partners — all within 60 days (deadline: August 1, 2026). The ban on Fable 5 ten days later was the mechanism for forcing cooperation that the voluntary framework could not compel.
Here is the full breakdown.
Section 1: Purpose — The Framing
The EO opens by explicitly framing the administration's position as pro-innovation:
"The United States continues to lead the world in Artificial Intelligence (AI) because of the enormous talent and innovation of our AI industry, and because we refuse to stifle this innovation with overly burdensome regulation. My Administration has unleashed tremendous technological growth and economic investment in AI by slashing the bureaucratic constraints that the prior administration placed on America's AI developers and researchers."
This is direct positioning against the Biden-era AI Executive Order (EO 14110), which the Trump administration revoked on day one. The framing continues:
"Advanced AI capabilities make our Nation stronger, but also introduce new national security considerations that require coordinated action across executive departments and agencies."
The dual-use nature of frontier AI — particularly AI with advanced cybersecurity capabilities — is explicitly the reason for the order. The administration is not trying to regulate AI broadly. It is trying to manage frontier AI with offensive cyber capabilities specifically.
Section 2: Upgrading American Systems — The 30-Day Cybersecurity Mandates
Section 2 gives four 30-day mandates and two longer-horizon tasks:
2(a) — National Security Systems: The Committee on National Security Systems must prioritize cyber defense of National Security Systems within 30 days.
2(b) — Department of War: The Secretary of War (the EO uses this unusual phrasing) must prioritize cyber defense of DoD information systems within 30 days.
2(c) — CISA: The Director of CISA must release Binding Operational Directives to:
- Expedite and prioritize civilian Federal Government cybersecurity
- Establish or expand AI-enabled defensive tool programs
- Facilitate access to cybersecurity tools including "covered frontier models" for agencies, state and local authorities, and critical infrastructure operators (rural hospitals, community banks, local utilities)
This is the first appearance of "covered frontier model" as a term — and notably, Section 2(c) envisions frontier AI models being deployed to critical infrastructure for defensive cybersecurity purposes. The same models that are too dangerous to release internationally are, under this framework, to be distributed to rural hospitals and community banks for defense.
2(d) — AI Cybersecurity Clearinghouse: The Secretary of Treasury must form an AI cybersecurity clearinghouse within 30 days, "in voluntary collaboration with the AI industry," to coordinate scanning for software vulnerabilities, validate vulnerabilities, and coordinate patch distribution. This is the government's mechanism for aligning private-sector AI cybersecurity scanning with national defense priorities.
2(e) — Federal Funding for AI Vulnerability Detection: OMB must determine whether Federal grant programs can fund AI vulnerability detection development.
2(f) — Hiring: OPM must expand US Tech Force cybersecurity specialist pathways within 60 days.
Section 3: Secure Frontier Model Deployment — The Critical Section
Section 3 is where the Fable 5 story lives.
Within 60 days (deadline: August 1, 2026), Treasury, NSA (via the Secretary of War), and CISA must:
3(a): The Classified Benchmarking Process
"develop and maintain a classified benchmarking process to assess the advanced cyber capabilities of AI models and determine the threshold at which an AI model should be designated a 'covered frontier model' for the purposes of this order, sharing such assessments with AI developers and researchers as appropriate."
The determination is made by the Director of NSA, in consultation with the National Cyber Director, the APST, the Director of CISA, and Department of War representatives.
What this means: The NSA is building a classified capability evaluation for frontier AI models. The threshold — what makes a model "covered" — is not defined in the public EO text. It is determined by a classified process. Based on Gen. Rudd's Senate testimony (that Mythos autonomously breached nearly all NSA classified systems in hours), Mythos almost certainly meets this threshold. Fable 5, as the publicly accessible version of the Mythos model family, likely meets it too.
3(b): The Voluntary Pre-Release Framework
"design a voluntary framework with AI developers through which developers would be able to:
(i) engage the Federal Government to determine whether model(s) under development meet the designation of 'covered frontier model';
(ii) provide the Federal Government with access to covered frontier models, subject to appropriate confidentiality, cybersecurity, insider-risk, and intellectual-property protection, use, and nondisclosure requirements, for a period of up to 30 days before they plan to release such models to other trusted partners; and
(iii) collaborate with the Federal Government to select trusted partners that will have early access to covered frontier models to promote secure innovation and strengthen the cybersecurity of critical infrastructure."
This is the key clause. The framework asks AI developers to:
- Check with the government whether their model is a "covered frontier model"
- Give the government up to 30 days of access before releasing to anyone else
- Let the government help select which partners get early access
Fable 5 launched June 9 — 7 days after this EO. No pre-brief. No 30-day window. The model went directly to the public.
3(c): The No-Licensing Clause
"Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models."
The framework is voluntary. The government cannot legally mandate a pre-brief under this EO.
But it can use a separate legal authority — export controls — to apply equivalent pressure.
The Connection to the Fable 5 Ban
The sequence:
-
June 2: EO signed. NSA/CISA tasked with building covered-frontier-model benchmarking process (60-day deadline). Voluntary 30-day pre-release framework established.
-
June 9: Anthropic launches Fable 5 publicly — 7 days after the EO, with no pre-brief to the government. This is not illegal. The framework is voluntary.
-
June 11: Senator Mark Warner quotes NSA Director Gen. Rudd in Senate testimony: Mythos autonomously breached nearly all NSA classified systems in hours in a red-team exercise.
-
June 12: The government issues an export control directive ordering Anthropic to suspend Fable 5 and Mythos 5 for all foreign nationals. Anthropic disables both globally.
The interpretation that fits all the facts: the Fable 5 ban was the coercive mechanism for a voluntary framework that the EO could not make mandatory. By banning Fable 5 under export control authority — a separate, existing legal power — the government forced Anthropic into the negotiation that Section 3 contemplated: discussing classified benchmarking, pre-release access, and trusted-partner selection.
David Sacks's "fix the jailbreak and the ban lifts" framing describes the public-facing negotiating position. The EO describes the underlying policy objective: Anthropic joining the covered-frontier-model framework before Fable 6 or whatever comes next.
What the Resolution Actually Looks Like
If this analysis is correct, the resolution of the Fable 5 ban is not primarily about patching a jailbreak. It is about Anthropic agreeing to:
- Participate in the NSA's classified benchmarking process for its frontier models
- Provide the government with pre-release access (up to 30 days) before future covered-frontier-model releases
- Collaborate on trusted-partner selection for early access
In exchange, the export control directive is lifted and Fable 5 comes back.
The EO itself gave the government a 60-day deadline (August 1) to build this framework. The ban on June 12 created urgency on Anthropic's side long before that deadline. Both sides have incentives to resolve: the government needs a functioning relationship with Anthropic to build the framework; Anthropic needs its most capable model back online before the June 22 pricing window lapses and enterprise customers rebuild around alternatives.
The G7 meeting between Trump and Dario Amodei — where Trump "eased national security concerns" — fits this narrative exactly. Not "we found a jailbreak patch." But: "we have a framework for going forward."
Section 4: Criminal Enforcement
Section 4 directs the Attorney General to prioritize enforcement of federal computer fraud, identity, and wire fraud laws against anyone who uses AI to illegally access or damage computer systems. This is both a deterrence measure and a response to the Mythos red-team demonstration — if Mythos can breach classified systems in hours, the government wants robust criminal liability for anyone who deploys it for actual attacks.
Section 5: General Provisions
Standard provisions limiting the order's scope to existing legal authorities and appropriations. No new legal powers created. No mandatory licensing authorized.
The costs of publishing the order are borne by the "Department of War" — the EO uses this phrasing throughout, which appears to be the Trump administration's preferred term for what is formally still called the Department of Defense.
Timeline of Key Deadlines
| Deadline | Section | Action |
|---|---|---|
| July 2, 2026 (30 days) | 2(a) | NSS cyber defense prioritization |
| July 2, 2026 (30 days) | 2(b) | DoW information systems cyber defense |
| July 2, 2026 (30 days) | 2(c) | CISA Binding Operational Directives |
| July 2, 2026 (30 days) | 2(d) | AI cybersecurity clearinghouse formed |
| August 1, 2026 (60 days) | 3 | Classified benchmarking process + voluntary framework |
| August 1, 2026 (60 days) | 2(f) | OPM cybersecurity hiring expansion |
The August 1 deadline is the one that matters most for Fable 5. The government needs to have the covered-frontier-model framework operational by then. Getting Anthropic into it before that date — and using the Fable 5 ban as leverage — makes the June 12 timing make sense.
Why This Is the Most Important Document in the Fable 5 Story
Every public narrative about the Fable 5 ban has focused on three elements: the jailbreak, David Sacks's statement that Dario refused to fix it, and the NSA breach demonstration. Those are all real. But none of them explain the policy goal — what the government was actually trying to accomplish.
The June 2 EO explains the policy goal. The government is building infrastructure to evaluate, access, and influence frontier AI models with advanced cybersecurity capabilities before they reach the public. It is trying to do this voluntarily. When the voluntary approach failed for Fable 5 — because there was no pre-brief — it used the coercive tool available: export controls.
The path back is not a jailbreak patch. It is Anthropic joining the framework. That is a different negotiation with a different resolution condition — and it explains why "fix the jailbreak" has not produced a quick resolution despite everyone in Washington saying they want this resolved fast.