Anatomy of a Shakedown
Pressure converting into property: Washington flipped the kill switch on a frontier AI model while asking the industry to hand over a piece.
The ultimatum arrived by phone; the letter arrived at 5:21
On Friday, June 12, around one o’clock Eastern time, Anthropic got a call from the U.S. government. The message was blunt: take down your two most powerful AI models, and you have ninety minutes to do it (”How Amazon and the White House ended Anthropic’s Fable,” Axios, https://www.axios.com/2026/06/13/anthropic-amazon-white-house). A source close to the company told Fortune there had been no prior warning of a national-security threat (”How a warning from Amazon led the White House to shut down Anthropic’s Mythos model,” Fortune, https://fortune.com/2026/06/14/how-a-warning-from-amazon-led-the-white-house-to-shut-down-anthropics-mythos-model/). By 5:21 p.m., Commerce Secretary Howard Lutnick had delivered a formal export-control directive to Anthropic CEO Dario Amodei — a letter the company says did not spell out the specific concern (”Anthropic disables Fable and Mythos AI models after U.S. government bars it from giving foreigners access,” Fortune, https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/).
The order barred any foreign national — anywhere, including Anthropic’s own non-citizen employees inside the United States — from accessing Claude Fable 5 and the more powerful Mythos 5 it is built on (Fortune, https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/). Unable to filter foreign nationals out of a live product in real time, Anthropic did the only thing the order left it room to do: it switched both models off for everyone (”Amazon’s Jassy Alerted White House to Anthropic Fable 5 Security Flaws, Triggering Export Ban,” MLQ News, https://mlq.ai/news/amazons-jassy-alerted-white-house-to-anthropic-fable-5-security-flaws-triggering-export-ban/). Its other models, including Opus 4.8, stayed online (Fortune, https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/).
That is not a routine event. Washington has long used export controls to keep advanced chips out of rival hands, but it had never before reached past the hardware and switched off an American AI model itself (”Anthropic Pulls Its Most Powerful AI Models After U.S. Bars Foreign Access,” TIME, https://time.com/article/2026/06/13/anthropic-fable-mythos-ban-US-security/). The most capable system Anthropic had ever shipped to the public went dark three days after launch — and eleven days after the company confidentially filed to go public (”Will Anthropic’s Fable 5 Be Back? Answers,” TECHSY, https://techsy.io/en/blog/anthropic-fable-5-suspended).
Set that against what the president had been saying the week before, and you have the sequence our infographic scores.
The sequence the scorecard maps
The timeline runs in five beats. Equity talk: on June 5, aboard Air Force One, the president said there were “concepts where pieces could be given to the American public,” who would “essentially become a partner” with AI companies (”US Exploring Government Partnerships with AI Firms, Trump Says,” Bloomberg, https://www.bloomberg.com/news/articles/2026-06-05/us-exploring-government-partnerships-with-ai-firms-trump-says). Model launch: on June 9, Anthropic shipped Fable 5, its first publicly available Mythos-class model (”Anthropic releases Claude Fable 5, its first Mythos-class model for the public,” The Next Web, https://thenextweb.com/news/anthropic-claude-fable-5-mythos-public-release-ipo). Stake talks: on June 10, in the Oval Office, the president said he would soon meet roughly 12 to 15 top executives to discuss “giving back” to the public (”Trump says he thinks AI companies will agree to ‘giving back’ to the public,” Reuters via Investing.com, https://www.investing.com/news/stock-market-news/trump-says-he-thinks-ai-companies-will-agree-to-giving-back-to-the-public-4735788). Export directive: on June 12, the Commerce Department’s order landed (Fortune, https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/). Global shutdown: within hours, the models were off (TIME, https://time.com/article/2026/06/13/anthropic-fable-mythos-ban-US-security/).
Extortion, as a legal matter, has elements — discrete things a prosecutor must show. Our scorecard borrows that structure not to allege a crime, but to be honest about which parts of this story the public record already supports and which it does not. Four chips are green. Two are amber. Two are red. The art argues guilt; the chips concede the proof gap. That tension is the design.
Coercive pressure is green: a global shutdown imposed by directive on a 90-minute clock, sweeping in allied nations and the company’s own staff, is coercive on its face (Axios, https://www.axios.com/2026/06/13/anthropic-amazon-white-house; Fortune, https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/). Property of value is green: AI-company equity is worth hundreds of billions — a recent round valued Anthropic near $965 billion — and the administration was openly interested in a piece of it (Fortune, https://fortune.com/2026/06/13/anthropic-disables-fable-mythos-export-controls-national-security-threat/; Bloomberg, https://www.bloomberg.com/news/articles/2026-06-05/us-exploring-government-partnerships-with-ai-firms-trump-says). The demand is green, with a caveat we will not bury: the “pieces / partnership / give back” language is real and on the record, but it was a public, industry-wide demand, and the concrete equity talks were reported to center on OpenAI, not Anthropic (”The Trump administration might take an equity stake in OpenAI,” TechCrunch, https://techcrunch.com/2026/06/06/the-trump-administration-might-take-an-equity-stake-in-openai/). There is no public evidence that Anthropic was privately told to hand over a stake. We score the demand as real; we do not score it as a private squeeze on this company. Compelled action is green: Anthropic changed its conduct and disabled the models to comply (MLQ News, https://mlq.ai/news/amazons-jassy-alerted-white-house-to-anthropic-fable-5-security-flaws-triggering-export-ban/).
Then the chips turn amber and red, and this is where honesty earns its keep. Intent is circumstantial: the timing is suggestive, but a documented security narrative blocks certainty (Fortune, https://fortune.com/2026/06/14/how-a-warning-from-amazon-led-the-white-house-to-shut-down-anthropics-mythos-model/). The link between the equity push and the export directive is contested and open — connected, on the current record, by proximity rather than proof; the administration’s AI adviser David Sacks publicly denied any connection to the companies’ earlier disputes (Fortune, https://fortune.com/2026/06/14/how-a-warning-from-amazon-led-the-white-house-to-shut-down-anthropics-mythos-model/). Wrongfulness is the same: was this a proportionate national-security action, or pretext dressed as security? We do not know, because the one party that could settle it has not shown its work.
The one fact the innocent story strains to absorb
If you want a single test that separates “security” from “leverage,” it is proportionality. The trigger, by the government’s own account relayed through reporting, was a jailbreak: Amazon — Anthropic’s largest investor, with roughly $13 billion committed — told the administration its researchers had used prompts to pull restricted cybersecurity information out of Fable 5, and CEO Andy Jassy carried the finding to Treasury Secretary Scott Bessent (”Anthropic’s Top AI Scientist Locked Out of His Own Company’s Most Powerful AI Because He’s Not American,” IBTimes UK, https://www.ibtimes.co.uk/ai-expert-andrej-karpathy-anthropic-tech-regulation-1802715). Sacks said a “highly credible trusted partner” had identified the jailbreak and that Amodei refused to fix it before the order issued (Fortune, https://fortune.com/2026/06/14/how-a-warning-from-amazon-led-the-white-house-to-shut-down-anthropics-mythos-model/).
Anthropic’s answer is pointed: the technique exposed only a small number of previously known, minor vulnerabilities, reproducible on other public models including OpenAI’s GPT-5.5, and not unique uplift from Mythos (”The U.S. government imposes export controls on Anthropic’s Fable 5 and Mythos 5,” KuCoin, https://www.kucoin.com/news/flash/us-government-imposes-export-controls-on-anthropic-s-fable-5-and-mythos-5). And it is not only Anthropic saying so. The company shared the Amazon report with Katie Moussouris, CEO of Luta Security, who told Axios the government’s response seemed far out of line with what the research actually showed — the researchers had found vulnerabilities by asking the very questions a defender would ask, which is exactly what the model was built to do (Axios, https://www.axios.com/2026/06/13/anthropic-amazon-white-house).
Even granting the government every benefit of the doubt, the gap between “a narrow, reproducible jailbreak” and “switch the entire model off worldwide, including for allies and your own employees, on ninety minutes’ notice” is enormous. Narrow problems usually get narrow fixes: a patch, a safeguard, a vetting layer for the high-capability tier. A global kill order is the heaviest instrument available, and it was reached for first. If the same vulnerability turns up in a competitor’s model and draws a patch window rather than a kill switch, the selective-treatment reading gets much stronger. That non-event is one of the most informative things to watch this week.
The counter-case is real, and we will not pretend otherwise
A responsible indictment includes the case for the defense. Deemed-export controls — the doctrine that giving a foreign national access to controlled technology counts as an “export,” even inside U.S. borders — are a normal, codified tool, not an invention for this occasion (”15 CFR 734.13 — Export,” eCFR, https://www.ecfr.gov/current/title-15/subtitle-B/chapter-VII/subchapter-C/part-734/section-734.13). Mythos genuinely raises cyber-risk concerns; Anthropic said as much at launch and built Fable 5 with safeguards that fall back to a weaker model on cybersecurity, biology, and chemistry queries (The Next Web, https://thenextweb.com/news/anthropic-claude-fable-5-mythos-public-release-ipo). The warning did not come from Amazon alone — at least five other companies flagged the model to senior officials in the same window (Axios, https://www.axios.com/2026/06/13/anthropic-amazon-white-house). Senior White House officials told Politico the export controls were a last resort after hours of trying to get Anthropic to cooperate, and an administration official argued other models were not treated as national-security threats because they did not clear the capability bar Mythos set (Fortune, https://fortune.com/2026/06/14/how-a-warning-from-amazon-led-the-white-house-to-shut-down-anthropics-mythos-model/; Axios, https://www.axios.com/2026/06/13/anthropic-amazon-white-house). Reporting also raised worry about adversary or China-linked access, a concern Anthropic disputes — none of that is fabricated, and a reader who waves it away is not being skeptical, but credulous in the other direction.
There is friction on the facts, too: Anthropic says it notified the government before the June 9 release and drew no objection, while an administration source counters that the company knew a jailbreak was possible and shipped anyway (Axios, https://www.axios.com/2026/06/13/anthropic-amazon-white-house). So the question is not whether there was pressure. There was. And it is not whether advanced AI can implicate national security. It can. The question is narrower and sharper: was national security the reason, or the cover?
What the week has added
As of Monday, June 15, the models are still dark. Anthropic is racing to reverse the controls and says it is working to restore access, but it has named no date and disclosed no mitigation checklist (”Anthropic Races to Reverse Fable 5, Mythos 5 Export Controls,” Yahoo Finance / BeInCrypto, https://finance.yahoo.com/sectors/technology/articles/anthropic-races-reverse-fable-5-033440732.html). The most likely outcome, by most reads, is “restored — with conditions,” and the conditions are exactly what the public is not being shown (TECHSY, https://techsy.io/en/blog/anthropic-fable-5-suspended). One person familiar told Axios the letter amounts to a de-facto licensing regime whose lasting effect is that companies “will not screw with the White House” (Axios, https://www.axios.com/2026/06/13/anthropic-amazon-white-house).
The markets have started pricing it. Pre-IPO Anthropic shares in thin synthetic secondary markets — a Hyperliquid perpetual that tracks where traders expect the equity to price — fell about 3.7% over the weekend to roughly $1,627, off post-launch highs above $1,800 (”Anthropic’s pre-IPO shares fall as US government shuts down Fable, Mythos models,” CoinDesk, https://www.coindesk.com/markets/2026/06/13/anthropic-s-pre-ipo-shares-fall-as-us-government-shuts-down-its-most-powerful-ai-model). The reaction is not all one way — some retail traders read the shutdown as proof the model is too good to leave in foreign hands (”Anthropic’s Fable 5 Ban Sparks Global Concern,” Stocktwits, https://stocktwits.com/news-articles/markets/equity/anthropic-s-fable-5-ban-sparks-global-concern-former-us-ai-adviser-questions-safety-practices-europe-revives-sovereign-ai-call/cZKfebdR7d1). Abroad, the European Union — which had only just negotiated Mythos access — used the moment to revive its call for technological sovereignty (Stocktwits, https://stocktwits.com/news-articles/markets/equity/anthropic-s-fable-5-ban-sparks-global-concern-former-us-ai-adviser-questions-safety-practices-europe-revives-sovereign-ai-call/cZKfebdR7d1). The shutdown is rippling outward exactly as a demonstration of power would.
The ask is disclosure, not a verdict
We are not claiming a crime was proven. We are claiming the public record now shows a coercion-shaped sequence in which the government demonstrated, in a single Friday, that it can switch off a frontier American product overnight on grounds it has chosen not to fully disclose — at the precise moment it was courting equity-like concessions from the same industry (Bloomberg, https://www.bloomberg.com/news/articles/2026-06-05/us-exploring-government-partnerships-with-ai-firms-trump-says; Reuters via Investing.com, https://www.investing.com/news/stock-market-news/trump-says-he-thinks-ai-companies-will-agree-to-giving-back-to-the-public-4735788).
There is a clean way to settle which story is true. Publish the directive. Cite the specific legal authority. Release a technical basis that matches the breadth of the remedy. Explain why a narrower fix was unavailable. If the security case is as serious as claimed, disclosure vindicates the government. If it cannot survive daylight, the public is owed that knowledge too. Anthropic, for its part, calls the whole thing a misunderstanding and says it is working to restore access (”Anthropic on X,” Anthropic, x.com/AnthropicAI/status/2065597531644743999).
Until those documents exist, the timing is doing the talking — and the timing reads, unmistakably, as pressure converting into property.