The Trump administration was close to allowing Anthropic to restore access to its Fable 5 model as of June 27, 2026, Reuters reported, ending what had been a two-week freeze triggered by a government directive barring foreign nationals from using both Fable 5 and Mythos 5.

The restriction originated with a directive issued by the Trump administration requiring Anthropic to block foreign national access to the two models, AP News reported on June 12. That same day, Anthropic confirmed it had begun working with the US government to restore access, according to the company's own statement. The two models — Fable 5 and Mythos 5 — sit at the frontier of Anthropic's current lineup, and a hard cutoff by nationality would have had immediate operational consequences for any enterprise or research customer running cross-border workloads against the API.

The path to restoration has been asymmetric. On June 26, the US government sent a letter to Anthropic co-founder Tom Brown addressing Mythos 5 access, but explicitly did not extend the same approval to Fable 5, CNBC reported. That left Fable 5 in a separate regulatory lane, one that, as of a day later, the administration appeared ready to unblock.

The mechanics of how Anthropic enforces nationality-based access controls are not detailed in public disclosures, but the requirement itself points squarely at the API layer — specifically, the kind of identity and jurisdictional verification that enterprise contracts typically handle through terms of service rather than real-time gating. Building hard enforcement at inference time, rather than contractually, is a non-trivial engineering and compliance lift. Whether the government's directive specifies a technical implementation standard, or leaves the method to Anthropic, has not been publicly confirmed.

The broader context here is the growing US government tendency to treat frontier AI models as export-sensitive assets — a framing that maps, somewhat imperfectly, onto existing ITAR and EAR frameworks developed for hardware and software with clear military applications. Applying those frameworks to general-purpose language models creates obvious ambiguities: the same model weights that can accelerate drug discovery can, in theory, assist in areas the government would prefer to gatekeep. That tension has been building for several years, and the Fable 5 / Mythos 5 episode is the sharpest commercial friction point it has produced so far.

Worth flagging: the split treatment of Mythos 5 and Fable 5 — one cleared, one still pending — is unusual. It suggests the government is making capability-differentiated judgments, not blanket model-tier rulings. If that is the operating principle, every future model release at the frontier tier may require its own individual clearance process, which would impose a structurally new compliance overhead on Anthropic and, by extension, on any AI lab releasing models at comparable capability levels.

For enterprise customers, the practical lesson is about concentration risk. Two weeks of restricted access to a flagship model, with no clear public timeline for resolution, is the kind of supply-chain disruption that procurement and architecture teams will now need to model explicitly — particularly those with international workforces or multi-jurisdiction deployments. Redundancy strategies that felt theoretical six months ago look considerably more concrete today.

Anthropic's original statement on access restoration efforts frames the company as an active partner with the government rather than a reluctant compliance subject — a positioning consistent with its longstanding emphasis on AI safety and its existing relationships with federal stakeholders. Whether that posture accelerated the Fable 5 resolution, or whether the timeline simply reflects the pace of internal government review, is not possible to determine from public information.