Uneven Restoration

Verdict: Claude Fable 5's restoration is not a comeback. It's a probation.

Seventeen days offline, June 12 to July 1, over what Commerce called national security concerns, and the model returns not with a clean bill but with a leash. 50% of weekly usage limits through July 7. KYC identity verification appearing on the signup flow. A sibling model, Mythos 5, still locked out of public hands entirely, restricted to a handful of government agencies while Fable gets to go home. That asymmetry is the whole story: one model rehabilitated for the masses, one kept in the room with the adults.

Read Commerce Secretary Lutnick's letter and the "restoration" stops looking like vindication. Anthropic didn't win an argument — it signed terms. Proactively detect and address security risks, coordinate on future model releases, report malicious activity: that's not a policy reversal, that's a supervised-release agreement dressed up as a press win. The company gets to call this a return to normal. The fine print says otherwise.

Nobody involved wants to be precise about what actually happened. 9to5Mac flags that Anthropic "has not confirmed whether the rollout applies to all users worldwide at once or gradually" — after nearly three weeks of downtime, the company still can't commit to a straight answer about who gets the model back and when. That vagueness is doing work. It lets Anthropic frame this as resolved while quietly staggering exposure, watching for whatever triggered the ban in the first place.

The tell is the usage cap. If this were truly cleared — no more national security concern, no more export license required — there's no reason to throttle returning users to half their normal allowance for a week. That's not a technical rollout constraint. That's a company hedging against its own model, in public, while calling it a restoration. Fable 5 is back. It is back on a curfew, wearing an ankle bracelet, standing next to a twin that isn't allowed out at all.

For nineteen days this summer, a silence rippled through every workflow built on top of Claude Fable 5 and Mythos 5 — not a crash, not a bug, but a government letter. On June 12, the Commerce Department's Bureau of Industry and Security told Anthropic to cut off "any foreign national, whether inside or outside the United States, including foreign national Anthropic employees" from the models, and overnight, tools that had quietly become load-bearing — coding agents, research pipelines, whatever people had wired into Claude Platform and Claude Code — went dark for a huge swath of their users. The interesting thing isn't that it happened. It's how fast the field metabolized it as normal, then treated the July 1 reversal as the actual news, when the reversal is the part that should worry people more.

Because what came back on July 1 wasn't symmetrical. Fable 5 returned to full daylight — general availability across Claude.ai, the Claude API, AWS, Google Cloud, and Microsoft Foundry — while Mythos 5 came back fenced, restricted to approved customers inside something called Project Glasswing. That's not a rollback to the prior state. That's the state choosing, in public, which model gets to be infrastructure and which gets to be classified. Commerce Secretary Howard Lutnick's language on the withdrawal — "the controls in the June 12 letter are withdrawn" — reads like a clean reset, but the two-tier outcome underneath it is the actual signal: one model gets treated like a utility, the other like an export-controlled asset with a name that sounds like a NATO codeword.

The trend worth naming isn't "AI got un-banned." It's that the field just watched, in real time, what happens when creative and technical infrastructure gets treated as a foreign-policy lever — and the answer is: it vibrates in sympathy with decisions that have nothing to do with craft. Nineteen days of downtime, confirmed end to end from the June 12 directive to the July 1 restoration, didn't just interrupt a service — it exposed how many people had stopped asking whether their tools could be switched off, because the switch had been invisible until it wasn't. Anthropic's own framing — that it would "begin restoring access" the day after the controls lifted — is corporate calm dressed over a genuinely destabilizing fact: creative infrastructure now has a foreign-national clause.

What should unsettle people isn't the ban or the lift, it's the precedent the asymmetry sets. If Mythos 5 stays gated behind an approval list while its sibling model goes fully open, the field has just been shown its own near future: frontier capability increasingly ring-fenced by policy, general-purpose capability increasingly treated as safe to release wide. That's not resilience. That's triage, and everyone building on these models just watched it happen live, then called it good news because the lights came back on.