Anthropic sends technical team to Washington to negotiate lifting AI export ban (2026)

Anthropic dispatched senior technical staff to Washington over the weekend of June 14–15 for face-to-face meetings with White House and Commerce Department officials, seeking to negotiate the lifting of export restrictions imposed on its two most powerful AI models. The talks are ongoing. No deal has been announced. 1

How the shutdown happened

Dario Amodei, CEO of Anthropic — Anthropic CEO Dario Amodei. 2

The sequence began on Thursday, June 12, when Amazon CEO Andy Jassy placed a late-night call to senior White House officials. Jassy told them that Amazon researchers had used a series of prompts to get Anthropic's newly released Fable 5 — a publicly available version of the Mythos-class model — to produce information about cyberattacks that was supposed to be restricted. 2 Amazon and five other companies testing the release found the same workaround — an exploit that, in the government's view, unlocked the full cyber capabilities of the more powerful Mythos 5 underlying model. 3

What followed was a disputed series of phone calls on Friday, June 13, between Anthropic CEO Dario Amodei and senior White House, Commerce, and Treasury officials. A source familiar with Anthropic told Fortune the company was given 90 minutes to pull its newest model and was provided no prior communication of a national security threat. An Anthropic source also denies that executives were hard to reach — a Trump administration account claimed Amodei was at a wellness retreat and could not be contacted; Anthropic says its executives were in touch with the White House within 15 minutes. 2

White House AI adviser David Sacks offered a different framing. In a post on X, Sacks said a highly credible partner had identified the jailbreak and that the administration asked Amodei to fix the issue or withdraw the model — and that Amodei refused, forcing the export control as a last resort. Sacks said the administration hopes Anthropic will remediate the issue so Fable 5 can return to general release quickly. 2

Semafor also reported, citing unnamed sources, that the U.S. government suspected a Chinese-linked group had already used the jailbreak Amazon discovered — though it was unclear what evidence supported the suspicion. Anthropic told Semafor the White House never raised Chinese access in its conversations with the company, and that Anthropic prohibits access from within China. 2

The administration's position: a "recklessness" charge

A senior Trump administration official told Fox Business that Anthropic's response to the concerns constituted "recklessness" — and that the interaction damaged the relationship to the point where the administration believes it can no longer trust the company to engage responsibly on safety issues.

The official said the government had made clear to Anthropic before the Fable 5 launch that it would be the first test case for the new executive order on AI guardrails. Multiple sources told Fox Business that Treasury Secretary Scott Bessent told Amodei directly during one call that he was making a bad decision. 3

A source close to Anthropic disputes that the company brushed off the concerns or refused to fix them. Anthropic said in its public statement that U.S. national security authorities had not identified specific concerns — only that the government believed it had become aware of a method of bypassing the model's safeguards.

The gap between the two accounts is now at the center of the D.C. negotiations. According to the Financial Times, the export controls on Fable and Mythos are raising broader doubts about how the U.S. government intends to police the most powerful AI systems going forward — and whether any future model crossing that capability threshold would need government approval before release. 4

Stakes: the IPO, global access, and future release policy

Anthropic CEO Dario Amodei, Chief Product Officer Mike Krieger, and Head of Communications Sasha de Marigny — Anthropic leadership: CEO Dario Amodei (left), CPO Mike Krieger (center), and Head of Communications Sasha de Marigny at the company's 2025 developer conference. 3

The timing is acute. Anthropic confidentially filed its S-1 with the SEC on June 1 — the export restriction lands directly on top of its IPO runway. Both models are offline for all users, not just foreign nationals, because the Commerce Department's directive prohibited Fable 5 and Mythos 5 from being distributed to "any foreign national inside or outside the United States" — a scope so wide, encompassing non-citizen Anthropic employees, that the company said it had no choice but to disable both globally.

If the D.C. talks succeed, the path back to general release requires Anthropic to either patch the specific exploit Amazon identified or produce a technical argument persuasive enough to satisfy the administration. The Axios report citing an unnamed administration official suggested that no other current model is seen as posing the same national security risk — meaning any future model that crosses Mythos-level capability would face the same pre-release scrutiny. 5

Internationally, the shutdown has reignited calls for what European politicians describe as "sovereign AI." French politician Bruno Retailleau said the episode should serve as a "wake-up call," and Canadian Prime Minister Mark Carney warned the incident exposed the risks of relying too heavily on U.S. AI providers. Former French PM Édouard Philippe said AI is now critical infrastructure — and infrastructure controlled by others is infrastructure that others can unplug. 2

For Anthropic's IPO story, the outcome of the Washington talks matters as much as any product benchmark. Whether the company emerges from the weekend with a clear remediation path — or with a damaged government relationship and an uncertain model timeline — will shape how institutional investors read its next S-1 filing.