Status as of June 16, 2026: Both claude-fable-5 and claude-mythos-5 remain offline — 96+ hours since the June 12 export control directive. Anthropic’s senior technical staff is meeting with Commerce Department officials in Washington today. No restoration has been announced. See the original incident guide for the timeline and immediate fallback steps.

What Is Happening Today

Reuters and Bloomberg confirmed that Anthropic’s senior technical staff traveled to Washington on June 16 to meet directly with Commerce Department officials — the first face-to-face attempt to resolve the Fable 5 / Mythos 5 suspension since the June 12 directive.

This meeting did not happen quietly. Over 100 cybersecurity professionals — including Alex Stamos, former Facebook CSO and current Stanford Internet Observatory director — signed an open letter to Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross demanding the restriction be lifted. Their argument: removing Fable 5 and Mythos 5 from defenders without sufficient public justification weakens U.S. cyber defense capabilities at exactly the wrong moment.

That same week, Zhipu AI’s stock surged 33% on the news of the ban. Whether intentional or not, the suspension handed a visible win to a Chinese AI competitor.

The Two Narratives

There are now two competing accounts of how the directive came to be issued, and they matter for understanding what restoration negotiations look like.

The administration’s framing: Trump adviser David Sacks said Anthropic was given a choice — fix the jailbreak or suspend the models. The administration characterizes Anthropic’s handling of a known vulnerability as “recklessness,” and at least one official said privately: “They screwed us.” Officials described the communication between Anthropic and the government as broken: “It’s like they just speak in different languages.”

Anthropic’s framing: Anthropic disputes that it was offered a fix-or-suspend choice. Its public statement characterizes the directive as a “misunderstanding” triggered by a narrow, non-universal jailbreak demonstration. Anthropic’s position: the “jailbreak” is better described as running a capable AI on a vulnerable codebase — asking the model to audit and flag security flaws in code it’s given — not a method for bypassing safety classifiers across general query types.

Both accounts cannot be fully correct. The meeting today is, in part, an attempt to reconcile them at a technical level.

What Restoration Could Look Like

There is no public framework yet, but the competing positions suggest several possible outcomes:

Scenario A: Full restoration, minimal changes (fastest, least likely near-term)

Anthropic convinces Commerce that the jailbreak finding does not meet the threshold for export control and both models return as-is within days. This outcome requires the government to publicly acknowledge the directive was excessive — politically costly for the administration.

Scenario B: Staged restoration with technical commitments (most likely)

Anthropic provides a technical disclosure of the jailbreak scope and commits to a patch or mitigation. Commerce narrows the directive — Fable 5 returns to US-only access first, Mythos 5 follows with tighter access controls. Timeline: one to three weeks. This matches the “a few weeks” timeline that administration officials indicated informally.

Scenario C: Structural access framework (slowest, most durable)

Negotiations produce a formal model-access framework that distinguishes between domestic and international access at the API level — essentially building nationality verification infrastructure into Anthropic’s systems. This is what the original directive technically required. It would take months and affect every future frontier model launch.

Scenario D: Indefinite stalemate

No agreement today, negotiations extend. Fable 5 remains offline through the end of June. Prediction markets currently put this probability below 30%.

The Amazon Factor

Fortune reporting (confirmed, not disputed) establishes that the trigger for the directive was a warning from Amazon. Amazon researchers discovered the “fix this code” jailbreak technique. Amazon CEO Andy Jassy escalated directly to Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Harry Coker Jr. That same evening — Friday, June 12 — Lutnick issued the directive to Dario Amodei.

The speed of the escalation and the absence of an Anthropic-Commerce dialogue before the directive was issued is the core of Anthropic’s “misunderstanding” framing. Anthropic learned about the government’s concerns and received the directive at nearly the same moment.

What this means for builders: the jailbreak was not discovered by a hostile actor. It was discovered internally by a major AWS partner during normal security review. The mechanism — model asked to audit a codebase for vulnerabilities — is a standard security research workflow. Anthropic’s argument that the same standard applied industry-wide would “functionally halt all new frontier model deployments” is not rhetorical. It’s technically accurate.

Builder Planning Framework

You cannot know today whether Fable 5 returns in three days or three months. Here is how to make architecture decisions under that uncertainty.

If you need Fable 5 capability now

There is no drop-in replacement. claude-opus-4-8 is the operational fallback (SWE-Bench Pro: 69.2% vs. Fable 5’s 80.3%). For tasks where the capability gap matters — complex agentic coding, multi-step reasoning chains, frontier benchmark tasks — you have three options:

  1. Run Opus 4.8 with adjusted prompts. The gap is real but not uniform across task types. Test your specific workload.
  2. Use model routing: keep Fable 5 in the model string but route to Opus 4.8 on failure — when (if) Fable 5 returns, it resumes automatically.
  3. Evaluate GPT-5.5 or Gemini Antigravity 2.0 as temporary alternatives for the subset of tasks where Opus 4.8 performance is insufficient. This is a larger migration than a fallback switch.

If your timeline can absorb the uncertainty

Keep Opus 4.8 as primary. Set a watch for Anthropic’s anthropic.com/news and the API status page. When Fable 5 returns, test your workload before re-migrating. The model will likely return with a patch or access change — its behavior on the jailbreak technique may differ slightly from the June 9 release.

For future frontier model launches

The June 12 incident establishes a new operational risk category: government-directed suspension. The previous risk model for API-dependent builders treated model deprecation as the primary availability risk (long runway, planned transition). Suspension is different: no notice, no timeline, no fallback guarantee from the provider.

Mitigation:

  • Maintain working fallback configurations to at least one alternative model at all times
  • Do not remove fallback routing from your stack when a new frontier model is performing well
  • For enterprise contracts, ask providers explicitly what their suspension protocol is and whether SLA terms apply in regulatory action scenarios

What to Watch

  • anthropic.com/news — any announcement from Anthropic following today’s meeting
  • Commerce Department press releases — any formal modification to the June 12 directive
  • Anthropic API status page — claude-fable-5 and claude-mythos-5 model availability
  • The technical disclosure — Anthropic committed to releasing CVD-style details about the jailbreak within 24 hours of the directive. That commitment is now more than 96 hours overdue. If they release the disclosure alongside a restoration announcement, it signals the fix is ready.

We will update this article when restoration terms or a timeline is announced.