If you've been waiting for Fable 5 to come back — and if you build anything on Anthropic's API, you have been — here's where things stand on Day 15.

Today, the US government cleared Anthropic to restore Mythos 5 for US organizations that operate and defend critical infrastructure. Energy companies. Hospital networks. Financial services. Telecom. The government's logic: the model is too useful for defending these systems to keep offline. These orgs are getting Mythos access back without any additional identity verification requirements.

Fable 5 — the consumer-accessible version of the same frontier capability, the model most developers, subscribers, and Claude Code users care about — is still offline for everyone else. Anthropic updated its privacy policy on June 21 with a new "Verification Data" category, effective July 8. The category covers what the policy calls a government-issued identity document, a photo or video of the user's face, and what it terms "facial geometry templates" — biometrics, processed by a third-party vendor called Persona Identities.

The read from analysts: that July 8 date is Anthropic's mechanism for bringing Fable 5 back for US citizens without waiting for the Commerce Department to formally lift the export control directive. If Anthropic can prove you're American, the "foreign national access" concern that triggered the ban gets addressed operationally, even if the directive stays on the books.

The most powerful AI model the company has built is coming back first — for power plants and hospital networks. The model you've been trying to use all month is asking you for your face.

What happened with Garbarino

The timing of this week's developments is partly explained by congressional politics.

At an Anthropic-sponsored "Fly Out Day" session earlier this month, the company demo'd Mythos to members of Congress. House Homeland Security Committee Chair Rep. Andrew Garbarino (R-NY) came away alarmed. His account to Punchbowl News: Anthropic "told the model to find a vulnerability in a bank and empty accounts, and then it went and did it." He said it "then could find this vulnerability and fix it."

Garbarino is now pushing for federal pre-release access to advanced AI models before they reach the public. The demo is being reported on June 27 as a factor in why the ban has been harder to lift than some expected.

A few things worth being precise about here. Security analysts note that Mythos-class models do find real vulnerabilities in banking and ACH flows — this is documented. The specific capability Garbarino described (finding a vulnerability in a bank system) is real. The part about "emptying accounts" at arbitrary live consumer banks is likely describing a controlled demo environment, not a live attack on actual accounts. That distinction tends to get compressed in political retellings. But politically, it doesn't matter: the congressman ran the demo, came away scared, and is now a factor in Washington's posture on the ban.

The irony — and it is a genuine irony — is that Mythos is the model built specifically for cybersecurity-grade capability. Fable 5 is the consumer-accessible version. The thing Garbarino saw demonstrated is exactly the use case the critical infrastructure orgs are now getting back. The thing general users and developers want is still offline.

Source spread

• NBC News — US government gives green light for limited re-release of Mythos 5 — builder. Clean reporting on the Mythos critical infrastructure restoration notice. Confirmed by multiple follow-on outlets.
• Punchbowl News — This Republican knows Mythos. He's scared. — safety. Primary source on the Garbarino demo account. DC insider press; Garbarino quoted directly.
• CIO — Anthropic's new privacy policy offers US consumers a way around the Fable ban — builder. Best analysis of the biometrics-as-restoration-mechanism read. Makes explicit the July 8 connection to the ban.
• Cybernews — Anthropic updates privacy policy, includes ID verification — skeptic. Covers the policy update itself; flags the scope of data collection and the Persona third-party relationship.
• CNBC — Anthropic's Mythos set off a cybersecurity 'hysteria.' Experts say the threat was already here. — skeptic. Best independent grounding on what Mythos-class models actually can and can't do in security contexts.

What's real. What deserves a side-eye.

What's real:

• Mythos 5 is coming back for critical infrastructure orgs. That's confirmed by NBC News and The Next Web, citing government notification to Anthropic.
• The July 8 privacy policy update is real, verifiable, and specific. Persona Identities is a real biometrics vendor. The data categories collected are described in precise terms in the policy.
• The Garbarino demo happened. His characterization of it is his characterization. The underlying capability — finding real vulnerabilities in banking infrastructure — is documented in research going back months.
• The analyst inference connecting July 8 to Fable 5's restoration path is logical, not confirmed. Anthropic hasn't said "biometric verification unlocks Fable 5." But the timing is not a coincidence and the mechanism fits.

What deserves a side-eye:

• "Facial geometry templates not stored by Anthropic" is technically true. They're stored by Persona Identities instead. That's a different thing from not stored at all. Persona's data retention policies are not Anthropic's policies, and you're giving your biometrics to a company you've probably never heard of.
• The Garbarino demo account compresses a controlled-environment vulnerability demo into something that sounds like a live attack on consumer banking. The capability is real; the framing is politically amplified. Both things can be true simultaneously.
• The biometrics path for Fable 5 helps US citizens. It does nothing for international users. If you're outside the US and have been waiting 15 days, July 8 is not your answer. There is no current public path for non-US general access to Fable 5.
• "Critical infrastructure" is a defined category for Mythos restoration. General API builders who don't operate power grids or hospital networks are in the same queue as individual subscribers.

What to do about it

If you use Claude.ai (free or paid subscriber):

• Fable 5 is still offline. Your current experience is whatever model Anthropic routes you to by default — Opus 4.8 or Fable 5's non-restricted capabilities where available.
• Starting July 8, you may be asked to verify your identity to access Fable 5 when it returns. This is optional until it isn't — meaning you can likely continue using Claude without completing verification, but Fable 5 specifically may require it.
• You're being asked to submit government ID and a live selfie to a company called Persona Identities. Read their privacy policy before you do.
• If you're outside the US: there is no current public timeline for international Fable 5 restoration. Plan accordingly.

If you're a builder on the Anthropic API:

• Fable 5 for general API use has no confirmed restoration date. Mid-July is a reasonable estimate, contingent on the ID verification rollout going smoothly.
• If you operate infrastructure in energy, healthcare, financial services, or telecom, check Anthropic's status page and your account dashboard — Mythos 5 restoration for your tier is happening now.
• The ID verification flow will add friction to onboarding new users on production apps that require Fable 5. Budget time for this in your architecture planning.
• International users on your platform: no access path currently exists. Factor that into any Fable 5-dependent product commitments you're making to customers.

Further reading

• NBC News — US government gives Anthropic green light for limited re-release of Mythos 5 — the critical infrastructure restoration announcement
• Punchbowl News — This Republican knows Mythos. He's scared. — Garbarino's demo account, primary source
• CIO — Anthropic's new privacy policy offers a way around the Fable ban — best analysis of the biometrics/restoration connection
• Cybernews — Anthropic updates privacy policy, includes ID verification — the policy change itself
• CNBC — Mythos set off cybersecurity 'hysteria.' Experts say the threat was already here. — grounds the Garbarino demo in what the model actually does