# US Restricts Anthropic Model Access Over Security Risk > Source: > Published: 2026-06-15 21:05:15.903880+00:00 # US Restricts Anthropic Model Access Over Security Risk Several news outlets report that the US government directed Anthropic to suspend foreign access to its most advanced models. According to Reuters, a copy of a Commerce Department letter seen by Reuters and reported by ITNews shows US Commerce Secretary Howard Lutnick ordered Anthropic to suspend exports of Fable 5 and Mythos 5 to foreign nationals worldwide. Reuters and other outlets report Anthropic said it would "abruptly disable" the models after receiving the directive. Reporting by Reuters and the Guardian states the government cited a potential method to "jailbreak" safeguards that could enable identification of software vulnerabilities; Anthropic says officials provided only "verbal evidence of a potential narrow, non-universal jailbreak," per Reuters. Coverage in The Washington Post, The New York Times, and others notes the move follows earlier tensions after Anthropic declined certain US military uses and a related supply-chain blacklist. ### What happened **Anthropic**, the San Francisco AI lab, disabled its most advanced models after receiving a US export-control directive restricting foreign access. Reuters reports the company received an order to suspend access to Fable 5 and Mythos 5 for all foreign nationals, and Reuters quotes Anthropic saying it would "abruptly disable" those models. According to ITNews and Reuters, a copy of a Commerce Department letter seen by Reuters shows US Commerce Secretary Howard Lutnick directed the company to stop exports worldwide. ### Technical details Per reporting by Reuters and the Guardian, the Department of Commerce flagged a national-security concern tied to a potential method of bypassing safeguards, commonly described in coverage as a "jailbreak" that could allow Fable 5 to assist in identifying software vulnerabilities. Reuters reports Anthropic's public statement that the government provided only "verbal evidence of a potential narrow, non-universal jailbreak." Anthropic has told reporters that the bypass the government cited found only "minor" security flaws that other publicly available models can also find, according to company statements summarized in Reuters and ITNews. ### Context and significance, Editorial analysis Industry observers have framed the directive as a substantial escalation of US export-control policy for frontier AI. For years, export controls emphasized chips and tooling; multiple outlets, including The Guardian and The Economist, place this action as a shift toward directly limiting access to deployed models. This change matters to practitioners because it alters how organizations evaluate geographic access, model deployment, and risk exposure for high-capability systems. ### Operational and stakeholder response Reporting from ITNews and Reuters says senior Anthropic technical staff have been meeting with Department of Commerce officials, and ITNews reports a scheduled in-person meeting in Washington. Multiple outlets, including ITNews and Reuters, note more than 80 cybersecurity executives and experts signed an open letter urging the administration to lift restrictions, with signatories reported to include leaders from firms such as Nvidia and Adobe, per ITNews coverage. ### Observed patterns in similar cases, Editorial analysis Companies facing export or regulatory restrictions commonly confront three practical challenges: maintaining global customer SLAs while complying with locality-based controls, certifying and reproducing technical evidence that regulators cite, and managing investor and partner uncertainty during rapid policy escalations. Practitioners should view this as part of a broader pattern where governments increasingly demand demonstrable, repeatable evidence for narrow technical risks before allowing cross-border access. ### What to watch - •Whether the Department of Commerce produces documented, reproducible evidence of the cited jailbreak beyond the "verbal evidence" Reuters reports. - •Any formal regulatory guidance or precedent set by the Commerce Department that defines what constitutes an export-control trigger for deployed models. - •Responses from cloud providers, enterprise customers, and non-US governments on operational workarounds, regional deployments, or legal challenges. ### Implications for model governance, Editorial analysis This episode underscores the growing intersection of model safety, security, and export policy. Organizations deploying high-capability models should expect increased scrutiny of demonstrable attack vectors, and industry-wide debates will likely intensify over how to balance technical mitigations, transparency, and cross-border access without relying on ad hoc, late-stage export directives. ### Reported factual sources Key reported facts above draw on Reuters, The Washington Post, The Guardian, ITNews Australia, and related coverage in The New York Times and Time, each cited in their respective paragraphs above. ## Scoring Rationale Direct government export controls on deployed frontier models are an industry-shaking policy development that affects deployment, governance, and international access for practitioners and vendors. 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)