Trump Administration Lifts Export Controls on Anthropic's Mythos and Fable Models

The Trump administration’s Commerce Department has rescinded licensing requirements for Anthropic’s Mythos 5 and Fable 5 models, ending months of regulatory friction over AI export controls. According to Wired AI, Commerce Secretary Howard Lutnick formalized the decision in a letter to Anthropic cofounder Tom Brown, clearing the way for unrestricted distribution of both models, which had previously been approved only for select companies and government agencies.

Anthropic’s Safety Concessions Unlock Export Approval

The license removal comes contingent on Anthropic meeting specific commitments outlined in Lutnick’s correspondence. According to Wired AI, the company agreed to proactively identify and mitigate security vulnerabilities in both model families, establish formal protocols with the US government on AI safety standards, and strengthen technical safeguards to minimize the number of successful jailbreak attempts. These concessions represent a notable shift from Anthropic’s earlier position—the company had originally maintained that eliminating all circumvention vectors was technically infeasible.

Wired AI reports that National Cyber Director Sean Cairncross and Lutnick jointly led the administration’s negotiation effort. The regulatory standoff had centered on concerns that users could exploit Mythos’s restricted capabilities, particularly those relevant to offensive cybersecurity operations, by bypassing Fable’s safety architectures.

Strategic Repositioning in Government Relations

The path to resolution involved more than technical modifications. Wired AI notes that Anthropic altered its negotiation strategy and personnel, replacing CEO Dario Amodei with cofounder Tom Brown in direct meetings with administration officials—a change that government contacts reportedly welcomed on personal grounds. This diplomatic recalibration, combined with Anthropic’s reframing of jailbreak reduction as an engineering priority rather than a philosophical impossibility, facilitated the breakthrough.

The company’s revised stance effectively met the administration halfway: rather than relying on theoretical arguments about the limits of alignment, Anthropic pivoted to demonstrating concrete commitments to risk reduction and government collaboration.

Why This Matters

This resolution establishes a template for managing powerful AI export policy without blanket restrictions. For Anthropic, the lifted controls restore commercial flexibility—Mythos and Fable can now reach international clients and developers without per-transaction licensing delays. For the Trump administration, the negotiated safeguards create enforceable checkpoints on model deployment without the friction of a blanket ban.

The precedent may influence how future administrations handle other frontier models. Teams evaluating Anthropic for production deployments will no longer face geopolitical licensing uncertainty. However, the agreement’s enforceability depends on Anthropic’s ability to deliver on jailbreak mitigation at scale—a technical challenge the company implicitly acknowledged by accepting the commitment. Independent security researchers should monitor whether the agreed-upon safeguards hold up under adversarial testing, since the deal’s legitimacy rests on measurable risk reduction, not regulatory theater.