Eighteen days. That’s how long it took Anthropic to go from “national security threat” to “unprecedented” government-industry cooperation. Fable 5 comes back Wednesday, worldwide, with a new jailbreak fix that blocks 99% of the vulnerability the government flagged. The other 1%? Already-patched flaws.
Anthropic’s Fable 5 full restoration was confirmed — closing the chapter TF has tracked since the 12 June export control directive. “We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5,” Anthropic said in a statement. “We’ll begin restoring access tomorrow, and will share an update soon.”
Commerce Secretary Howard Lutnick confirmed the decision on X the same evening. “Over the past two weeks, we have worked closely with Anthropic to analyse and approve Fable 5 to ensure alignment across the US Government and strengthen America’s leadership in AI.”
Fable 5 is available again to global users on Claude.ai, Claude Code, and via API starting Wednesday, 1 July. Additionally, it will be included for up to 50% of weekly usage limits through 7 July for Pro, Max, Team, and select enterprise plans — a temporary access boost welcoming users back after 18 days offline.
What’s Happening & Why It Matters
What Amazon Found — and What the New Safeguard Blocks
Anthropic’s Fable 5 full restoration clarifies a detail TF could not confirm at the time of the original suspension. CNN reported that the trusted partner who found the jailbreak was Amazon. Anthropic implemented a new safeguard specifically to address and block the vulnerabilities Amazon reported. That safeguard is the mechanism enabling restoration. Anthropic said it worked closely with the government to implement a new safeguard that blocks the jailbreak 99% of the time. In 1% of cases, Fable’s output only included previously discovered or already-patched security flaws — meaning even the residual exposure carries no genuine new risk.
By contrast, Anthropic was candid about the trade-off the new safeguard introduces. The tighter classifier could result in Fable 5 flagging more benign requests — a direct cost to ordinary users in exchange for closing the security gap. The Commerce Department’s own Center for AI Standards and Innovation independently tested both the model’s previous and new safeguards, according to Anthropic — providing government-side verification rather than relying solely on the company’s own assessment.
The 18-Day Timeline
Anthropic’s Fable 5 full restoration completes an 18-day sequence that TF has covered as it unfolded. As TF reported in its Fable 5 suspension article, the Commerce Department issued the original directive on 12 June, forcing Anthropic to disable both models worldwide within hours. As TF covered in its Mythos 5 restoration article, Lutnick partially restored Mythos 5 for 100+ critical infrastructure organisations on 26 June — while Fable 5 remained fully offline. The full Fable 5 restoration on 30 June represents the final and complete resolution: both models available, with Mythos 5’s critical-infrastructure list restrictions presumably folded into the restoration, though Anthropic has not detailed whether any residual access tiers remain.
The Bureau of Industry and Security’s own language confirms the resolution’s finality. “Bureau of Industry and Security’s evaluation of the diversion risks presented by Claude Mythos 5 and Claude Fable 5, the controls in the June 12 letter are withdrawn,” Lutnick wrote. That is a complete withdrawal — not a partial adjustment.
“Unprecedented” Cooperation
Anthropic’s Fable 5 full restoration was termed by the White House as a triumph of government-industry collaboration rather than a reversal under pressure. White House chief of staff Susie Wiles wrote on X that the government and private sector had “worked together in a way we have never seen before and the foundation of America First is unprecedented.” By contrast, the situation is uneasily alongside the substance of the episode — a government directive that disabled a commercially available product worldwide within hours, based on a jailbreak that Anthropic itself characterised at the time as exploiting “previously known, minor vulnerabilities” rather than enabling a sophisticated cyber breach.
The crackdown drew criticism from tech executives and investors throughout its 18-day duration — in part because it handed valuable development time to Chinese open-source AI labs. China’s Zhipu is closing in on the top US AI models, with Anthropic and OpenAI both held back during the relevant window. TF’s coverage throughout June documented the same theme repeatedly: every day Fable 5 is offline is a day competitors — both domestic and Chinese — operate without an equivalent restriction.
Not the Last Word on AI Governance
Anthropic’s Fable 5 full restoration resolves the specific episode without resolving the underlying structural question it exposed. The US government’s desired role in regulating and evaluating frontier AI models before release is still up in the air — creating an ad hoc regulatory environment for AI companies. OpenAI rolled out GPT-5.6 to only a small set of approved customers the same week, following a direct request from the US government — the pattern TF documented in its Mythos 5 restoration article. Neither company’s experience is seen by either party as “voluntary” in practice.
The Trump administration faces an August deadline under its earlier executive order to create standardised benchmarks for evaluating the security risks of new AI models. That benchmark framework — not the specific Fable 5 episode — is the mechanism that will determine whether frontier AI labs face a predictable, transparent review process going forward, or another round of ad hoc directives issued and withdrawn under commercial and political pressure.
TF Summary: What’s Next
Fable 5 access resumes globally from 1 July 2026 across Claude.ai, Claude Code, and API. Anthropic will re-enable Fable 5 access via Amazon Web Services, Google Cloud, and Microsoft Foundry “as soon as possible.” The 50% weekly usage boost runs through 7 July. The Trump administration’s August 2026 deadline for standardised frontier AI security benchmarks is the next major milestone in the policy area.
MY FORECAST: Anthropic’s Fable 5 full restoration will not be the last government-imposed suspension of a frontier AI model in 2026 — but it establishes the template for how such episodes resolve. A directive issued rapidly on national security grounds, a company response combining technical fixes with active Washington negotiation, and a resolution within roughly two to three weeks noted publicly as a collaborative success. By contrast, the more consequential outcome is what the episode teaches every other frontier lab. Google DeepMind, Meta AI, and xAI have a concrete precedent for exactly how much runway they have before a jailbreak report — even one addressing “minor” vulnerabilities — can trigger a worldwide shutdown. Every lab racing toward Mythos-class capability will build safeguards and Washington relationships proactively, rather than reactively negotiating during an active suspension. The August benchmark deadline is where that precedent either is formal policy or is an improvised, case-by-case system.
Related Stories
- Anthropic Pulls Claude Fable 5 — Four Days After Launch, on Government Order
- Mythos 5 Returns for Critical Infrastructure — Fable 5 Remains Suspended
- Anthropic Accuses Alibaba of 28.8 Million Fake Queries to Clone Claude