Semafor reports the White House's decision to impose export limits on Anthropic's most advanced models was driven in part by suspicions that a China-linked group had accessed Mythos. The Commerce Department directed Anthropic to block foreign use of Fable 5 and Mythos 5; Anthropic disabled access to both models within 90 minutes of the government order, according to CryptoBriefing and The Wall Street Journal. David Sacks, co-chair of the President's Council of Advisors on Science and Technology, posted a public account on X saying the government had warned Anthropic of a Fable 5 jailbreak and that CEO Dario Amodei declined to fix it. Anthropic disputes Sacks's framing and says the government did not raise Chinese access in its conversations around the jailbreak, per Semafor.
What Happened
The Commerce Department directed Anthropic to block foreign use of Fable 5 and Mythos 5, and Anthropic disabled worldwide access to both models within 90 minutes of the government order, according to CryptoBriefing and The Wall Street Journal. The Wall Street Journal reports that Commerce Secretary sent a letter to Anthropic CEO Dario Amodei triggering the action. The Los Angeles Times and The Wall Street Journal confirm the models were taken down after the order was received.
The Sacks Account and Amazon Angle
David Sacks, co-chair of the President's Council of Advisors on Science and Technology and the administration's former AI czar, posted a public account on X on June 13, 2026, laying out the government's version of events. Sacks said a trusted partner of both Anthropic and the U.S. government tested Fable 5 and found a jailbreak that bypassed consumer guardrails to access the cyber capabilities of the underlying Mythos model. He said the administration asked Amodei to fix the bypass or de-deploy the model, and that Amodei declined, calling the jailbreak "not serious." Semafor, citing a person familiar with the matter, reports that Amazon CEO Andy Jassy was in contact with the administration about the jailbreak and that Amazon flagged the issue to the government; Amazon declined to confirm specifics, with a spokesperson telling Semafor that governments often seek its counsel on security risks. Sacks stated the export control was issued "reluctantly" and that "the ball is in Anthropic's court."
Anthropic's Position
Anthropic publicly stated that the bypass is "narrow and non-universal" and that the same result can be produced on other public models including OpenAI's GPT-5.5. The company has said it disagrees that a narrow jailbreak should necessitate the recall of a model used by hundreds of millions of people. An Anthropic spokesperson told Semafor that the White House "didn't raise Chinese access to Mythos in its conversations around the Fable jailbreak," and that Anthropic blocks access to its products from inside China.
China Access Allegation
Semafor, citing a person familiar with the matter, reports that the White House acted partly over suspicion that a China-linked group had accessed Mythos, raising the prospect of model extraction via distillation. The White House has not issued a formal public confirmation naming a specific group. Anthropic has separately noted it blocks access from China, and disputes the government's characterization of the jailbreak as a serious threat. The New York Times and WSJ document the broader national-security framing: the Commerce Department cited the risk that advanced cyber capabilities of Fable and Mythos would be available to parties outside the trusted access pool.
Context and Prior History
This is not the first access incident involving Mythos. Tom's Hardware reports that in April, unauthorized third parties reached the restricted model using data from a breach. Anthropic also lobbied previously for Mythos to be regulated as a cyberweapon - a position Sacks cited as inconsistent with refusing to pull the model over a jailbreak. Anthropic is separately suing the Pentagon following an impasse over autonomous weapons use of its models, per Tom's Hardware.
What to Watch
- •Whether a formal public finding names a specific China-linked group and documents confirmed access; current reporting credits Semafor with the allegation, with no formal government confirmation.
- •Technical evidence for model extraction or distillation; current reporting documents government concern but no confirmed instance of successful reverse engineering.
- •Whether Anthropic patches the Fable 5 jailbreak and whether the Commerce Department lifts restrictions after a fix, as Sacks indicated is the government's preferred outcome.
- •Broader precedent: this is the first known case of the U.S. government ordering a commercial AI company to disable consumer access to a frontier model on national-security grounds.
Bottom Line
An unprecedented U.S. policy action forced the worldwide shutdown of Anthropic's top two models within 90 minutes, combining a disputed jailbreak, an unconfirmed China-access allegation, and a public dispute between the administration and a major AI lab over model safety responsibility.
Scoring Rationale #
This is an unprecedented U.S. government intervention: the first known case of export controls forcing a commercial AI lab to disable consumer access to frontier models worldwide within 90 minutes. The story combines confirmed operational impact, a disputed jailbreak account with named government sources, an unconfirmed China-access allegation, and a public conflict between the administration and Anthropic that has direct implications for how all major labs manage model deployment, access vetting, and safety response going forward.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.