After the June 12 directive forced Anthropic to take its Mythos and Fable 5 models offline for every customer, not just the foreign nationals named, two weeks of negotiations have produced no resolution.
For two weeks, the AI export-control standoff between the Trump administration and Anthropic has produced no public resolution. After the Commerce Department's June 12 directive pulled Anthropic's newest models offline, the company complied broadly, suspending access for every customer rather than carving out only the foreign nationals the order named. Fourteen days later, the models are still gone, the company is silent, and the legal mechanism behind the order looks more durable than any narrow compromise the two sides might strike.
The mechanism is the story. Commerce invoked the Export Control Reform Act of 2018, which rests on the broader authority of the Export Administration Regulations (EAR), a regulatory framework first written under the 1979 Export Administration Act. According to a Just Security legal explainer, that statute was designed to control dual-use technologies from machine tools to encryption to semiconductors, not frontier language models. The June 12 directive puts those authorities to work against a deployed commercial AI model in a way that has no clear precedent, and Anthropic's own statement frames the order as touching every foreign national, including foreign-national Anthropic employees, anywhere in the world.
What Anthropic contests is not the statute but the premise. In its public statement, the company said it disagreed with the administration's view that 'a narrow jailbreak finding should recall a model deployed to hundreds of millions of users,' and that it had chosen to suspend access for all customers rather than attempt a narrow compliance carve-out for foreign nationals. The Verge's reporting on the standoff characterizes the action as an industry-wide escalation, while Just Security and CSIS frame it as a regulatory directive, not a criminal enforcement action. The administration is leaning on Commerce's administrative authority rather than the Justice Department to compel compliance.
The legal vehicle matters because it travels. The Export Administration Regulations can be applied to any 'emerging' technology the Commerce Department determines warrants control. According to CSIS's analysis, the precedent opens a path to similar directives against any frontier lab whose models meet a comparable capability profile. Anthropic is not the only US AI developer whose latest releases can plausibly solve cyberoffense problems or accelerate weapons-relevant research, and the same statutory reach would in principle extend to other developers with comparable capability.
Just Security, in its legal explainer, walks through how EAR item classifications and Foreign Direct Product rules extend jurisdiction to foreign-national employees of US companies, which is how a US company can find its own workforce inside the scope of a US export ban. That extraterritorial reach is what makes the directive feel different from prior AI safety interventions: it does not ask Anthropic to retrain or filter its model, it asks the company to wall off its own engineers from the system they built.
In the meantime, the diplomatic channel is narrow but active. Reuters reported that Anthropic and US officials met on Monday to resolve the dispute, indicating the parties have not abandoned negotiation, but no outcome has been disclosed and Anthropic declined to comment this week. The Verge's running account of the standoff notes that any resolution would still leave the precedent intact: even if the models return to customers next week, the Commerce Department would have established that a 1979-era trade statute is a usable instrument for AI containment.
What to watch is whether the precedent gets tested first by an expansion, not a retraction. If the directive's logic applies to Mythos-class capability profiles, the next move is whether the same Commerce authorities reach other US frontier AI developers. If they do, the negotiation inside Anthropic stops being a customer-access story and becomes the founding case of US frontier-AI export policy.