Anthropic announced late Tuesday that the US Department of Commerce had lifted export controls on its Claude Fable 5 and Mythos 5 models, ending an 18-day standoff since the June 12 directive. Commerce Secretary Howard Lutnick confirmed on X that the Bureau of Industry and Security had withdrawn the controls, stating “a license is no longer required for the export, reexport, or in-country transfer of the Mythos or Fable models.” Anthropic will begin restoring global access to Fable 5 on Wednesday, available for up to 50% of weekly usage limits through July 7 for Pro, Max, Team, and selected Enterprise users. Mythos 5, partially restored on June 26 for US organisations operating critical infrastructure, continues under government-approved conditions. YourNewsClub identifies the Lutnick statement as the most significant political element of the resolution: a Commerce Secretary publicly “approving” a specific AI model and endorsing it as strengthening “America’s leadership in AI” describes a model-by-model government review process that did not exist before June 12.
The 18-day standoff produced specific changes inside Anthropic. The company worked with the government to implement a safeguard that blocks the relevant jailbreak technique 99% of the time – a targeted fix rather than an architectural redesign. Anthropic co-founder Tom Brown led negotiations, replacing CEO Dario Amodei after Amodei became a focal point for administration criticism of AI safety advocates who supported the previous administration.
The resolution comes against a backdrop of concurrent tensions. In March, Anthropic sued the Pentagon after it labelled the company a “supply chain risk” for refusing to permit Claude’s use in autonomous weapons without human oversight. Defense Secretary Pete Hegseth signed with OpenAI instead. A federal judge subsequently called the restrictions “Orwellian.” The 18-day restriction also handed Chinese open-source developers uninterrupted access to global researchers who could not access Fable 5 during that window. YourNewsClub maps that competitive cost as the most commercially concrete consequence of the standoff beyond direct user disruption.
OpenAI rolled out GPT-5.6 to only a small set of approved customers last week following a government request, before any broader release. That restricted deployment and the Anthropic standoff together describe an emerging pattern: the US government is now operating an informal, ad hoc approval process for frontier AI models that lacks statutory authority, defined timelines, or publicly stated criteria.
Jessica Larn, who studies macro-level technology policy and infrastructure impact of AI, draws the governance question: “An August executive order deadline requires standardised benchmarks for evaluating AI model security risks. Without those benchmarks, the process that produced the June 12 letter could repeat at any moment with any model. The resolution of this specific incident does not resolve the absence of a legal framework.” Alex Reinhardt, who tracks financial systems and settlement infrastructure through digital protocols, places the commercial exposure: “An export control directive that can shut down global model access based on an informal assessment, with no prior notice or defined review period, is a material business risk that Anthropic’s IPO prospectus will need to address fully.”
YourNewsClub calls the August benchmark deadline the next consequential date in the Anthropic-government relationship, since it will determine whether the ad hoc approval dynamic of the past 18 days formalises into a structured review process or continues as informal case-by-case intervention that companies must navigate without predictable rules.
OpenAI rolled out GPT-5.6 to only a small set of approved customers last week following a government request. Two simultaneous government pre-release reviews for frontier AI models in the same week describe exactly the pattern that the August benchmark deadline is supposed to formalise.
Anthropic filed confidentially and targets a listing before the end of 2026. Carrying an active Pentagon legal conflict alongside the resolved – but precedent-setting – Commerce restriction as background facts in its prospectus will require legal framing that goes well beyond Tuesday’s announcement. Your News Club notes the prospectus disclosure obligation as the document that will most definitively characterise how Anthropic and its underwriters represent the governance risk that the past 18 days made visible.