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

The Trump administration has removed export controls on Anthropic's Claude Mythos 5 and Claude Fable 5 AI models, ending a standoff that had lasted nearly three weeks and effectively blocked global access to two of the most capable AI systems publicly associated with any American company. Anthropic confirmed it received notice from the Department of Commerce on Tuesday evening and said it would begin restoring access to both models on Wednesday, July 1.

Commerce Secretary Howard Lutnick announced the reversal, stating that the Bureau of Industry and Security had withdrawn the export controls that had been imposed on June 12. In a post on X, Lutnick said his office had worked closely with Anthropic over the preceding two weeks to analyze and approve Fable 5 and ensure alignment across the US government. A license, he confirmed, is no longer required for the export, reexport, or in‑country transfer of either the Mythos or Fable models.

How the Restrictions Started

The original directive, issued on June 12, required foreign entities including Anthropic's own employees working outside the United States to obtain export licenses before accessing either model. Complying with that requirement was practically unworkable at the scale Anthropic operates, so the company chose to disable both models entirely for all users rather than risk violation. The move blindsided much of the AI industry, since this marked the first time the US government had imposed export controls on a domestic AI company's consumer‑facing products.

The government's stated rationale was national security and cybersecurity risk. Officials had informed Anthropic that a method existed to circumvent the safety guardrails built into the models, raising concerns about potential misuse. Fable 5 had launched in early June as a publicly accessible version of Anthropic's Mythos class of models, designed with additional security layers to limit the most sensitive capabilities. Mythos 5, the more powerful underlying model, had already been operating under restricted access since April, available only to a carefully screened group of companies and federal agencies.

What Changed in the Negotiations

After Anthropic disabled both models, the company's co‑founder Tom Brown took the lead in negotiations with the Trump administration, a notable shift since CEO Dario Amodei had been handling government relations previously. The switch in lead negotiator was widely interpreted as a deliberate move to reduce friction, given that Amodei had been a vocal supporter of Kamala Harris during the 2024 presidential election and had been publicly outspoken on AI safety in ways that had drawn the administration's attention.

By late June, the Commerce Department had granted partial relief, allowing Mythos 5 access to a select group of approved companies and federal agencies. Sources had indicated around that time that a full restoration of Fable 5 was imminent.

The full reversal comes with conditions. Anthropic has agreed to proactively detect and address security risks in the models, work with the US government on protocols and standards for Mythos, Fable, and future model releases, and notify the government of any malicious activity it identifies. The company also confirmed it had implemented a new safeguard specifically targeting the jailbreak technique that had originally concerned officials. According to Anthropic, this new safeguard blocks the relevant exploit in 93 percent of cases, with the remaining instances producing output that only involves previously discovered or already patched security vulnerabilities rather than novel attack capability.

Anthropic noted that the new safeguard may result in the model flagging a higher number of benign requests as well, a tradeoff the company appears to have accepted as part of the agreement. Commerce's Center for AI Standards and Innovation independently tested both the previous and new safeguards before the controls were lifted.

The Competitive Pressure Behind the Reversal

The restrictions drew criticism well beyond Anthropic's own user base. A number of technology executives and investors raised concerns that the export ban was handing valuable time to Chinese AI developers who had been rapidly closing the capability gap with leading US models. Several Chinese open‑source models had gained significant traction during the weeks of restriction, with some approaching Mythos‑class performance at substantially lower cost.

That competitive framing appears to have been central to the resolution. White House chief of staff Susie Wiles described the government‑private sector coordination as unprecedented, framing the outcome in terms of strengthening American leadership in AI rather than limiting risk. The administration is now facing a separate deadline in August, set under a recent executive order, to establish standardized benchmarks for evaluating the security risks of new frontier models before release.

The episode has exposed a regulatory gap that the US government is still working to close. There is currently no formal, consistent process governing how frontier AI models are evaluated and approved before reaching the public, which created the ad hoc situation that produced the June 12 directive and the weeks of uncertainty that followed. OpenAI separately agreed around the same time to allow the administration to screen users of its newest model, a signal that government involvement in frontier AI deployment decisions is becoming a structural feature of the market rather than a one‑off intervention.

Broader Industry Implications

For Anthropic, the restoration ends a period of significant operational disruption and reputational uncertainty. The company's products had effectively become unavailable to a large share of its global user base, including enterprise customers and developers building on its API, at a moment when the AI market is moving quickly and switching costs between platforms are still relatively low.

The incident has also established a new precedent. Before June 12, the idea that a domestic AI company's consumer product could be placed under export controls overnight would have seemed far‑fetched to most industry observers. It is now precedent, and every major AI lab operating at the frontier now carries a category of regulatory risk that did not meaningfully exist six weeks ago. How companies price that risk into product strategy and government relations going forward will likely be one of the defining questions for the frontier AI sector through the rest of 2026.