Anamika Dey, editor
By TechSun News Desk | techsunnews.com | July 3, 2026 | Tech / AI | 5 min read 🚨
Nineteen days. That’s how long Anthropic’s most powerful AI models sat completely offline — shut down by a government order that gave the company just 90 minutes to comply.
On July 1, 2026, at 3:31 PM ET, that ended. Claude Fable 5 came back online for users worldwide. The US Department of Commerce lifted its export controls the evening before, and Anthropic spent the night preparing the rollout.
But the return came with conditions. And those conditions tell you a lot about where the relationship between Washington and America’s AI companies is heading.
If you followed our coverage when Anthropic was given 90 minutes to shut down its best models back in June — this is the rest of that story.
What Happened in Those 19 Days
The shutdown began on June 12, triggered by a report from Amazon researchers describing a jailbreak technique that had bypassed one of Fable 5’s safeguards. The method prompted the model to identify software vulnerabilities — and in one case, produce code showing how a vulnerability could be exploited.
The US government moved fast. The Commerce Department issued an export control directive barring any foreign national — including Anthropic’s own employees — from accessing the models. With no way to verify user nationality in real time, Anthropic pulled both Fable 5 and Mythos 5 offline for everyone, worldwide.
What followed was three weeks of quiet negotiations in Washington. Anthropic’s co-founder Tom Brown led the company’s discussions with federal officials. On June 26, the government gave partial clearance — Mythos 5 could return for a select group of US organisations working on critical infrastructure. Fable 5 stayed down.
Four more days of talks. Then on June 30, Commerce Secretary Howard Lutnick made the call. Export controls lifted. (Source: CNBC)
What Anthropic Had to Agree To
The return wasn’t unconditional. In his letter to Anthropic, Lutnick set out three requirements the company had to accept before access could be restored.
First: Anthropic must proactively detect and address security risks in its models — before they become public problems, not after.
Second: The company must work with the US government to help develop standards for how future frontier AI models are reviewed and approved before launch.
Third: Anthropic must report any malicious activity it finds within its models directly to the government.
These aren’t suggestions. They’re the price of getting Fable 5 back. (Source: Al Jazeera)
Anthropic also retrained its safety classifier to block the original jailbreak technique. The company says the fix works in more than 99% of attempts — though it acknowledged that the stronger filter increases false positives, meaning some legitimate coding and debugging queries now get rerouted to a less powerful model. (Source: The Next Web)
Was the Shutdown Justified?
That depends on who you ask — and the answer matters, because this won’t be the last time it happens.
Anthropic consistently pushed back on the severity of the original finding. The company said the jailbreak technique did not expose any unique Mythos-level cybersecurity capabilities and described it as a borderline case for Fable 5’s guardrails, not a critical vulnerability. Several independent cybersecurity experts agreed, calling the threat significantly overblown.
The US government took a harder line. And given that we covered how the Trump AI crackdown is already giving China an opening, the political pressure to reverse the decision quickly was real. Every day Fable 5 sat offline, Chinese open-source competitors kept shipping.
Anthropic itself called out the broader risk directly: the shutdown handed valuable time to Chinese developers working to close the capability gap. That argument landed.
What Access Looks Like Right Now
If you’re a regular Claude user, Fable 5 is available again — but with a catch you need to know about before July 7.
Until July 7, 2026: Pro, Max, Team, and select Enterprise subscribers can use Fable 5 for up to 50% of their weekly usage limit at no extra cost. After that, Fable 5 moves to usage credits. It will no longer be included in standard subscription plans.
Standard Enterprise subscribers: Fable 5 has never been included in the base seat price. All usage is billed through credits — that hasn’t changed.
Cloud platforms: Anthropic is restoring access on AWS, Google Cloud, and Microsoft Foundry as quickly as possible — though full availability there is still rolling out.
Mythos 5: Still not publicly available. It remains limited to approved US organisations through Project Glasswing — the government’s trusted partner framework for frontier AI access. (Source: VentureBeat)
The Bigger Question This Leaves Open
Fable 5 is back. But the episode changed something that won’t quietly go back to normal.
Before June 12, AI model releases were essentially a company decision. You built it, you tested it, you launched it. What happened to Anthropic — and separately to OpenAI’s GPT-5.6 launch, which was restricted to trusted partners at the government’s request the same week — signals that frontier model releases are no longer purely commercial events. They’re negotiated deployments.
Anthropic put it plainly in its redeployment post: government involvement in AI releases requires a durable, transparent process with clear rules, applied equally across all frontier model developers. The company called for that process to be codified in law.
Whether Washington moves that direction — or keeps handling each model case by case — is the question the entire AI industry is now watching.
| 🟡 EDITOR’S OBSERVATION
The most telling detail in this whole story isn’t the shutdown or the return. It’s the timeline: 90 minutes to comply on June 12, 19 days to negotiate a way back. Anthropic had no emergency protocol, no regulatory playbook, no precedent to follow. Neither did the government. Both sides were improvising. The three conditions Lutnick attached to the restoration are essentially the beginning of that playbook being written in real time — and every AI lab in the US is now reading it carefully. |
💬 WE WANT TO HEAR FROM YOU
| Do you think the US government was right to shut down Fable 5 — or was it an overreaction?
A) Right call — AI this powerful needs government oversight B) Overreaction — it handed China a free advantage for 19 days C) The real problem is there are no clear rules yet for any of this Tell us in the comments — this one’s going to keep happening. |
❓ FREQUENTLY ASKED QUESTIONS
| Q: Why was Fable 5 shut down in the first place?
Amazon researchers found a jailbreak technique that bypassed one of Fable 5’s safeguards, allowing the model to identify software vulnerabilities and, in one case, produce code showing how a vulnerability could be exploited. Amazon CEO Andy Jassy reportedly flagged this to the White House. The Commerce Department responded with an export control directive on June 12 — barring all foreign nationals from accessing the models and giving Anthropic just 90 minutes to comply. With no way to verify user nationality in real time, Anthropic pulled both Fable 5 and Mythos 5 offline globally. |
| Q: Can I use Fable 5 for free now that it’s back?
Not exactly. Until July 7, 2026, Pro, Max, Team, and select Enterprise subscribers can access Fable 5 for up to 50% of their weekly usage limit at no extra cost as part of a temporary welcome-back offer. After July 7, Fable 5 requires usage credits — it moves out of standard subscription plans. Standard Enterprise seat subscribers have always been billed for Fable 5 through credits and nothing changes there. If you want to try Fable 5, July 7 is your deadline for the included access window. |
| Q: What is Mythos 5 and why isn’t it fully back yet?
Mythos 5 is the underlying cybersecurity-focused model that Fable 5 is built on top of. While Fable 5 is designed for general public use with safety guardrails, Mythos 5 carries the full cybersecurity capability stack — which is precisely why it triggered more concern. The government has cleared Mythos 5 only for approved US organisations working on critical infrastructure, accessed through a framework called Project Glasswing. There is no public timeline for broader Mythos 5 availability. |
Disclaimer: This article draws from reporting by CNBC, Al Jazeera, VentureBeat, The Next Web, 9to5Mac, Axios, CIO, and MarketScale. All information reflects what was available as of July 3, 2026.
