Skip to content
News Incident

Anthropic sent senior technical staff to Washington to resolve the Mythos export-control dispute after Amazon CEO Andy Jassy triggered the shutdown

· by Pondero Newsdesk

The short version

Anthropic technical staff held virtual and in-person meetings with White House officials on June 14, 2026, to resolve export controls that took Fable 5 and Mythos 5 offline. Fortune separately confirmed Amazon CEO Andy Jassy warned the administration after Amazon researchers demonstrated a way to extract restricted Mythos outputs.

Anthropic sent senior technical staff to Washington to resolve the Mythos export-control dispute after Amazon CEO Andy Jassy triggered the shutdown

Anthropic flew senior technical staff to Washington and opened a round of virtual and in-person meetings with White House officials on June 14, 2026, in an effort to end the export-control crisis that has kept Fable 5 and its underlying Mythos 5 model offline since June 12. A separate Fortune report published the same day named Amazon CEO Andy Jassy as the person whose warning to senior administration officials set the shutdown in motion.

What happened

The immediate trigger, per Fortune, was a demonstration by Amazon researchers who used a series of prompts to extract Mythos-class model outputs related to cyberattacks in ways that Fable 5's safeguards were supposed to block. Jassy passed those findings to senior administration officials before the Commerce Department acted.

Whether Amazon conducted the tests on its own initiative or at the government's request remained unclear at publication time. An Amazon spokesperson told Fortune: "As a leading cloud provider that serves a large number of private and public sector customers, it's not uncommon for governments to seek our counsel on potential security risks. When they occur, we don't share the details of these discussions."

Anthropic received the Commerce Department directive at 5:21 pm Eastern Time on June 12, per Anthropic's own blog post, and said the letter did not provide specific details of the government's national security concern. The company was reportedly given 90 minutes to pull both models. Because the directive barred distribution to any foreign national, including non-citizen employees inside the US, Anthropic concluded it had no choice but to disable the models for all users. Its less powerful Claude models, including Claude Opus 4.8, were not affected.

On June 14, Axios reported that Anthropic's technical staff had held virtual meetings with White House officials since the administration's initial outreach the previous Friday, and that sources on both sides said they were eager to resolve the matter. Administration officials had previously claimed Anthropic had not engaged seriously, a characterization the Axios source disputed.

Why it matters

The shutdown is the first time the US government has applied national security export controls to block access to a commercial AI model already in broad public use, per Fortune's June 13 reporting. The scale is significant: the controls affected not just users outside the US but also non-citizens working inside the country, a category that includes a substantial portion of the AI industry's workforce.

For Anthropic specifically, the timing is acute. The company confidentially filed for a public listing earlier in June at a reported valuation of $965 billion, per Fortune. A prolonged shutdown of its most capable models, or any signal that government action could routinely disrupt model access, creates a material risk question for prospective investors.

The episode also sharpened a long-running political fight. The Trump administration designated Anthropic a "supply chain risk" for Pentagon contractors in early March after the company declined contract terms allowing its models to be used for "all lawful purposes," citing concerns about autonomous weapons and mass domestic surveillance. Anthropic is challenging that designation in court. White House AI adviser David Sacks, who has repeatedly clashed with Anthropic publicly, offered his own account of the Fable 5 dispute over the weekend: he said a trusted partner of both Anthropic and the government had identified a jailbreak, that Amodei was asked to fix the issue or withdraw the model, and that Amodei refused. Anthropic has not confirmed that framing.

The reaction outside the US was swift. Former French Prime Minister Edouard Philippe called AI "critical infrastructure" that others can "unplug," per Euronews. UK MP Al Carns noted that British hospitals and researchers had been using Fable 5 before it went dark. Canadian Prime Minister Mark Carney cited the episode as evidence of the risks of depending on US AI providers.

Context and reactions

Anthropic's public position is that the jailbreak Amazon discovered was narrow, unlocking Mythos cybersecurity capabilities in one specific configuration rather than defeating all of Fable 5's safeguards broadly. The company also argued, in its June 12 blog post, that the same technique could elicit comparable outputs from other publicly available models including OpenAI's GPT-5.5, none of which faced similar controls.

AI policy analyst Dean Ball, who briefly served in the Trump administration, wrote on X that he could not determine whether the action amounted to "lawfare against Anthropic in particular or extreme national-security hawkery," calling it "simply cartoonish." Ben Murphy of the Institute for Progress warned the directive could discourage AI labs from being transparent with the government about model capabilities in the future.

Semafor reported separately, citing unnamed sources, that the US government suspected a Chinese-linked group had already used the jailbreak Amazon found. Anthropic told Semafor that the White House had not raised Chinese access in its conversations with the company, and that Anthropic prohibits access to its products from within China. The basis for the government's suspicion was not confirmed.

Treasury Secretary Scott Bessent told Anthropic CEO Dario Amodei directly during one call that he was making a bad decision, per Politico. White House officials also told Politico the export controls were a last resort after hours of requests to Anthropic to cooperate.

What to watch next

Two outcomes are most consequential. First, whether the DC meetings produce a partial-restoration order, possibly limiting Fable 5 access to US citizens only, or whether Anthropic can demonstrate a remediation patch that satisfies the government's technical objection. Second, whether Amazon's internal vulnerability demonstration becomes part of any public record, a court filing in Anthropic's pending litigation over its Pentagon supply-chain designation, or remains in government channels only.

A longer-term question is whether an administration official's statement to Axios, that any future model crossing Mythos's capability threshold would need to go through the government before release, becomes formal policy or a one-off warning.

Sources