Models offline
Fable 5 + Mythos 5
Directive issued
June 12, 2026
DC meetings began
June 13–14
IPO target
October 2026
Anthropic flies staff to DC to fix the White House fight over Fable 5 and Mythos 5
Senior Anthropic technical staff flew to Washington on June 14 to meet face-to-face with White House officials after a Commerce Department export control order forced the company to pull Fable 5 and Mythos 5 globally. Both sides say they want a deal, but the White House publicly accused Anthropic of not engaging seriously. The outcome will shape the precedent for whether any narrow AI jailbreak is sufficient grounds for a full deployment shutdown — and it lands 16 weeks before Anthropic's target IPO date.
Anthropic dispatched senior technical staff to Washington on Sunday to meet face-to-face with White House officials, two days after a Commerce Department export control order forced the company to shut off Fable 5 and Mythos 5 for every customer worldwide. 1
Both sides say they want to resolve the standoff quickly. But they are not yet in the same room on what a fix looks like.
How the ban started
The immediate trigger was a phone call. According to The Wall Street Journal, Amazon CEO Andy Jassy personally called Treasury Secretary Scott Bessent and other senior officials to report that Amazon researchers had used Fable 5 to extract "information that could be used in cyberattacks." 2
That call preceded the Commerce Department's Bureau of Industry and Security issuing an export control directive on Friday, June 12, ordering Anthropic to block all foreign nationals — including those in allied countries and Anthropic's own foreign-born employees — from Fable 5 and Mythos 5. 3
David Sacks, co-chair of the President's Council of Advisors on Science and Technology, said on X that the White House had asked Anthropic CEO Dario Amodei the previous week to patch or de-deploy Fable 5 — and that Amodei declined. 4
Anthropic complied with the directive but challenged the reasoning. In a public statement, the company said the government cited a "narrow, non-universal jailbreak" comparable to vulnerabilities in other deployed models, including OpenAI's GPT-5.5, and warned that applying this standard consistently would "essentially halt all new model deployments for all frontier model providers." 5
The DC trip
By Friday afternoon Anthropic's technical staff had started holding virtual meetings with White House officials, according to a source close to the company. Over the weekend, those discussions went in-person: senior technical staff flew to Washington to meet with administration officials directly. 1
The White House, however, framed the situation differently. Administration officials told reporters that Anthropic had not engaged in a "serious manner" up to that point — a public pressure tactic that Anthropic did not immediately dispute. The Wall Street Journal confirmed the trip separately, describing Anthropic as "racing to resolve" the situation. 6
The source told Axios that both sides say they are eager to reach a deal.
What Anthropic needs to fix
The technical question is whether Anthropic can demonstrate that the Fable 5 jailbreak has been patched, or sufficiently contained, before the White House will lift the export restriction. The harder question is the precedent: if Anthropic accepts the government's legal theory that a narrow jailbreak justifies an immediate global shutdown, every frontier model released in the US faces the same exposure.
Anthropic's public posture on this has been consistent: the jailbreak is real but minor, it also exists in models from competitors that have not been restricted, and the government's response is disproportionate to the actual risk. 5 The company needs a deal that addresses the specific jailbreak without validating the broader standard.
What's at stake for the company
The timing compounds the pressure. Anthropic filed a confidential S-1 with the SEC on June 1 and has named Goldman Sachs, Morgan Stanley, and JPMorgan Chase as lead underwriters for an IPO targeting October 2026. The global model shutdown creates a direct revenue hole — enterprise customers who cannot use the models are already evaluating alternatives — and raises questions for prospective IPO investors about the stability of Anthropic's government relationship and the reliability of its product availability. 7
The ban also created a secondary problem: many of Anthropic's own foreign-born engineers could no longer access the models they built.
統計カードを読み込んでいます…
Outside the US, the reaction was sharper. The European Commission said it was assessing the "practical consequences" of the ban and stated that contingency measures "should not be discriminatory against partners." British MP Tom Tugendhat said the global shutdown cut off dozens of UK hospitals and researchers overnight. French politicians including Jordan Bardella and Bruno Retailleau called it a "wake-up call" for Europe to back domestic alternatives like Mistral AI. 3
Bloomberg called the episode a "warning to Silicon Valley" — the first time the US government has pulled a deployed frontier AI model offline and the clearest demonstration yet that the Trump administration is prepared to act unilaterally on AI security concerns regardless of commercial fallout. 7
The talks are continuing. No timeline for a resolution has been disclosed.
参考ソース
- 1Axios — Anthropic flies staff to D.C. to clean up White House fight
- 2WSJ — Amazon CEO's talks with U.S. officials triggered crackdown on Anthropic models
- 3SiliconAngle — White House forces Anthropic to disable new frontier models
- 4Bloomberg Law — Anthropic's Fable curbed by US over jailbreak issue, Sacks says
- 5Anthropic — Fable and Mythos access
- 6WSJ — Anthropic dispatches staff to D.C., racing to resolve AI export restrictions
- 7Bloomberg — Anthropic block marks US reversal, warning to Silicon Valley
このコンテンツについて、さらに観点や背景を補足しましょう。