Commerce Secretary Howard Lutnick called it two weeks of 'alignment' talks; reporting ties the original block to Amazon CEO Andy Jassy's White House outreach after Amazon researchers flagged specific prompt risks — exposing how fast an AI model's national security verdict can be…
The 19 days between "national security risk" and "approved for global release" tell you more about how US AI governance actually works than any policy paper.
On June 12, 2026, the Commerce Department ordered Anthropic to pull its Claude Fable 5 and Claude Mythos 5 AI models offline, with reporting indicating the company had roughly 90 minutes to comply. By July 1, both models were cleared: Mythos 5 went out to more than 100 US institutions the previous Friday, and Fable 5 is now restored worldwide for Claude subscribers and enterprise customers.
The reversal wasn't technical. No new safety review, no patched vulnerability, and no revised threat assessment was published to justify the flip. Commerce Secretary Howard Lutnick framed it instead as a two-week exercise in ensuring the model was "aligned" with the US government. Anthropic's announcement echoed the language, thanking the Department for lifting export controls on Fable 5 and Mythos 5.
But reporting from The Wall Street Journal, relayed by 9to5Mac, ties the original block to a different lever: Amazon CEO Andy Jassy's outreach to the White House, after Amazon researchers raised concerns that specific prompt sequences could coax Fable 5 into producing information useful for cyberattacks. A frontier AI model was deemed a national-security export risk on June 12 and cleared for global release on July 1, on a timeline that began when a hyperscaler customer with cloud compute leverage picked up the phone.
The shape of the order matters as much as the outcome. The June 12 directive barred any foreign national, including Anthropic's own non-US staff, from accessing the models. That goes well beyond the usual export-control logic, which targets cross-border shipments. It also constrains domestic access by a workforce the company has already vetted. The 90-minute compliance window was set for the same reason as the broad scope. The order wasn't trying to be calibrated. It was trying to be felt.
What followed was the diplomatic part. Anthropic spent two weeks negotiating with Commerce. The publicly visible result is a model that had been limited to "select trusted organizations" via a program called Project Glassing, a tightly gated preview for Mythos 5, and is now available globally on Claude.ai, Claude Code, Claude Cowork, and the Claude Platform. For Pro, Max, Team, and select Enterprise customers, Fable 5 usage is capped at 50% of weekly limits through July 7, after which Anthropic moves to usage credits rather than throttling. The model itself didn't change. The terms of access did.
The Lutnick framing is the part worth sitting with. The Secretary described the process as analyzing Fable 5 to ensure "US Government alignment," not as a re-evaluation of the underlying cybersecurity concern that reportedly triggered the block. That is a different category of judgment. One is a technical risk assessment. The other is a political settlement about whose interests the model serves. Commerce can claim both, but they answer different questions, and they can move in different directions on the same calendar.
That distinction matters because the next time a frontier AI model lands in the same gap, capable enough to raise security concerns and central enough to the AI economy that pulling it would be commercially disruptive, the precedent now points at the second track. A two-week commercial-diplomacy window, opened by the right phone call from the right hyperscaler, is a faster mechanism than any technical re-review the government currently runs. Whether that is good or bad depends on whether you trust the phone-call mechanism more than the technical-review mechanism. For now, the trade press is treating the answer as obvious. It works if you are big enough to make the call.
The unresolved piece is Mythos 5. Wired reports both models were cleared under the same order, but only Mythos 5 was rolled out to institutions last week, and Anthropic's announcement focuses on Fable 5. If the same commercial-diplomacy channel that brought Fable 5 back is what eventually opens Mythos 5 to a wider audience, the watch item is who makes the next call to the White House, and which customer they represent.