# US Export Controls Disable Anthropic Models Globally > Source: > Published: 2026-06-17 08:24:03.558113+00:00 # US Export Controls Disable Anthropic Models Globally The U.S. government issued an export-control directive on June 12, 2026 requiring suspension of access to Fable 5 and Mythos 5 for any foreign national, and Anthropic disabled both models worldwide within 90 minutes to comply, per its public statement and Reuters. The order cited national security concerns over a reported jailbreak of Fable 5's guardrails. Per Tom's Hardware, David Sacks (co-chair of the President's Council of Advisors on Science and Technology) said the administration had warned Anthropic in advance and that CEO Dario Amodei declined to patch the vulnerability or pull the model before controls were imposed. Anthropic disputes this, calling the jailbreak narrow and non-universal and noting comparable results are possible on other public models including GPT-5.5. Fortune and other outlets report the action has accelerated calls for sovereign AI across Europe and allied countries. ### What happened Per Anthropic's public statement, the U.S. government issued an export-control directive on June 12, 2026 requiring suspension of access to Fable 5 and Mythos 5 for any foreign national, whether inside or outside the United States; Anthropic stated it received the directive at 5:21pm ET and 'must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.' Reuters, Forbes and Fortune reported the order cited national security authorities and referenced a reported technique for bypassing Fable 5's guardrails. Multiple outlets note Anthropic kept other Claude models online while the two models were disabled worldwide. (Anthropic statement; Reuters; Forbes) ### Pre-history and contested accounts Per Tom's Hardware (reporting on statements by David Sacks, co-chair of the President's Council of Advisors on Science and Technology), the U.S. government warned Anthropic before the directive that a trusted partner had discovered a jailbreak in Fable 5 - specifically, a bypass of the guardrails separating the consumer model from Mythos's unrestricted cybersecurity capabilities. Sacks said the administration asked CEO Dario Amodei to patch the bypass or pull the model, that Amodei declined, and that the government issued the export control 'reluctantly' after that refusal. Per Semafor (cited in Tom's Hardware), Amazon flagged the jailbreak to the administration, with Amazon CEO Andy Jassy in direct contact; Amazon did not confirm the details on the record. WSJ also reported Amazon CEO's conversations with U.S. officials triggered the crackdown. (Tom's Hardware; WSJ; Semafor via Tom's Hardware) Anthropic disputes this account. In public statements, Anthropic says the jailbreak is narrow and non-universal - per Tom's Hardware, Anthropic's position is that it 'amounts to asking the model to read a codebase and identify software flaws' - and that the same result can be achieved on OpenAI's GPT-5.5 and other public models. Anthropic disagrees that a narrow bypass justifies disabling a model used by hundreds of millions of people, and says it is working to restore access once the jailbreak is addressed. (Anthropic statement; Tom's Hardware) ### Technical context Industry-pattern observations: Public reporting and commentary highlight two technical frictions. First, nationality-based export controls assume reliable identity enforcement; The Conversation and other coverage observe that cloud providers typically infer user location rather than citizenship, and that location controls can be evaded by VPNs or other routing techniques. Second, the trigger reported by outlets was a demonstrated jailbreak technique that allegedly bypassed safety filters on Fable 5, a pattern observers note is a known class of risk when models expose advanced capabilities without robustly tamper-resistant guardrails. (The Conversation; Anthropic statement; Reuters) ### Context and significance Coverage in Fortune and Reuters frames the action as a policy precedent: a U.S. export-control order used to curtail foreign access to frontier commercial models. Reporting shows this moment renewed political pressure in Europe for sovereign AI and prompted national debate in allied countries over reliance on U.S.-hosted models. Analysts quoted in multiple outlets point to the broader geopolitical strain between policy, commercial AI deployment, and the technical limits of access control. (Fortune; Reuters) ### What to watch Observers should track three categories of signals. First, regulatory follow-ups or clarifications from U.S. agencies specifying scope and technical criteria for export controls. Second, product and partnership changes by model vendors - for example, localized hosting, vetted partner programs, or adjustments to access-control tooling. Third, government and procurement responses in Australia, Europe and other allies addressing onshore compute, data residency, and open-source or domestically hosted models. (Fortune; The Conversation; Reuters) ### Limits of reporting What is reported publicly does not include a detailed technical forensic report from the U.S. government or a verbatim, independently published demonstration of the alleged bypass technique. Anthropic's statement says the government did not provide specific details of its national security concern. The contested accounts from Sacks and Anthropic represent two sides of a dispute neither party has fully resolved in public. (Anthropic statement; Reuters; Tom's Hardware) ### Implications for teams For practitioners: document dependency maps showing which workflows rely on hosted frontier models, consider fallback options (different models, on-premise or open-source stacks, vetted partner programs), and follow vendor notices closely for changes to SLAs, partner access, and compliance regimes. This episode is likely to accelerate policy conversations and procurement due diligence even where direct technical fixes are still developing. ## Scoring Rationale This story establishes a high-impact regulatory precedent - a U.S. export-control order that removed global access to frontier models within 90 minutes, with direct operational consequences for teams relying on hosted AI. The contested accounts between the administration and Anthropic, the Amazon-trigger angle, and the accelerated sovereign-AI debate across allied nations all add substantive policy and geopolitical weight beyond the initial access disruption. Practice interview problems based on real data 1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with. [Try 250 free problems](/problems)