Anthropic Sent a Lobbying Team to Washington Tuesday to Fight the Mythos Export Ban. The White House and Anthropic Still Disagree on Why It Happened.

Anthropic sent representatives to Washington on Tuesday to contest the export controls restricting its Mythos frontier model, according to Yellow.com. The goal: reverse or narrow the restrictions before they harden into standing precedent for how the U.S. government can constrain frontier AI deployments going forward.

Two Stories, One Decision

The White House and Anthropic are not telling the same story about why this happened.

According to Semafor, which reported the competing accounts on June 13, the administration imposed controls on Mythos partly because a China-linked organization had accessed the model. The concern: if Beijing had working access to Mythos, it could attempt to distill and replicate the model, and could exploit its vulnerability-detection capabilities against U.S. systems. Semafor's source said it's unclear which organization accessed the model or how.

David Sacks, a Trump adviser, posted a different framing on X. Sacks said the government received a warning that Fable 5, the consumer version of Mythos, could be jailbroken. When the administration notified Anthropic, Sacks alleged, CEO Dario Amodei told them the jailbreak was NOT a serious risk and declined to fix it. Sacks said the export controls were then imposed reluctantly.

A person close to the White House told Semafor that Amazon CEO Andy Jassy had been in contact with administration officials about the jailbreak, and that Amazon was the source of the original warning.

Anthropic's official response, relayed to Semafor through a spokesperson: the White House never raised Chinese access to Mythos in any of their conversations around the restrictions.

Three separate explanations. One action.

What the Controls Actually Did

The White House directed Anthropic to limit access to Mythos and Fable 5 to U.S. citizens only, which would have barred even Anthropic's own foreign-national employees from using the models. Anthropic responded by pulling both models from the market entirely, according to Semafor.

Yellow.com notes the decision was not routed through a standard rulemaking process. No formal regulatory guidance was issued. Politico, as cited by Yellow.com, reported that industry advocates viewed the move as a breach of the administration's earlier commitment to minimal AI oversight. Fortune characterized the result as a de facto licensing regime, one that gives the White House effective veto power over which frontier models can reach international users, without explicit statutory authority to do so.

That last point is the one that should worry every other frontier AI lab. If this precedent sticks, the government does not need a formal rule to ground a model. It needs a phone call.

The Strongest Case for the White House

It would be a mistake to dismiss the administration's security concerns as purely pretextual. Anthropic itself publicly stated that Mythos performs at a sophisticated level in finding software vulnerabilities, and flagged that capability as a potential public danger, according to Yellow.com. The company's own caution about Mythos's offensive security potential is a matter of record.

If a China-linked entity did access Mythos, the national security case for rapid action is real. Distillation — using a frontier model's outputs to train a copy — is a documented technique, and a model capable of identifying previously unknown software vulnerabilities would be extraordinarily valuable to a state-sponsored hacking operation. The administration's argument that speed mattered more than regulatory process has a logic to it, even if the process question remains legitimate.

The Factionalism Factor

The Verge's Tina Nguyen, writing June 17, adds a layer the other outlets don't address directly: White House internal factionalism. Nguyen, who covered Trump's first term, draws on documented patterns of competing administration factions leaking conflicting narratives to shape the media environment and protect their own positions. Her analysis suggests the multiple, contradictory accounts of the Mythos decision — from the jailbreak story to the Chinese access story to the process dispute — may reflect different factions within the administration offering self-serving framings rather than a single coherent government rationale.

Axios sources added yet another dimension, telling that outlet the administration simply did not like Anthropic's ideological approach: "Anthropic has not done a great job at trying to speak to the administration and appreciate the ideological differences," one source familiar with the administration's thinking told Axios. "It's like they just speak in different languages."

That framing doesn't tell us which account is accurate. It does explain why there are multiple of them.

What Comes Next

European governments, according to Yellow.com, have responded to the Mythos news by accelerating sovereign AI programs. Whether that development strengthens or weakens Anthropic's case in Washington is genuinely unclear: it could argue that export controls are pushing AI capability development abroad, or the administration could argue it confirms U.S. competitors are trying to close the gap and tighter controls are justified.

Anthropic's Tuesday Washington meetings are an attempt to resolve that before the export controls calcify. The unresolved question is whether the administration will clarify which of its own stated justifications — Chinese access, jailbreak risk, or both — actually drove the decision. Until it does, Anthropic's lobbying team is negotiating against a rationale that keeps shifting.