Dario Amodei and senior Commerce Department officials sat down in Washington D.C. today in the first documented face-to-face attempt to resolve the export controls that have kept Fable 5 and Mythos 5 offline since June 12.
What happened: On Thursday June 12, Amazon CEO Andy Jassy personally alerted senior administration officials that Amazon researchers had used a series of prompts to extract restricted cyberattack information from the Mythos-class model. The Commerce Department issued the export control directive the same day — giving Anthropic 90 minutes to comply. Amodei disputed the characterization, calling it a narrow, non-universal bypass rather than a full jailbreak. He refused to withdraw the model. Commerce acted.
The meeting: Today's session involved Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross on the government side. Anthropic was represented by CEO Dario Amodei, co-founder and Chief Compute Officer Tom Brown, and policy chief Sarah Heck. The government is expected to provide documentation of the alleged jailbreak — the first concrete evidence Anthropic says it has yet to receive. Whether the meeting produces a resolution, a technical remediation path, or a stalemate remains unclear.
The bigger picture — three converging pressure points:
- G7 summit (Evian, France): Amodei is scheduled to join an AI luncheon Wednesday alongside OpenAI's Sam Altman and Mistral's Arthur Mensch. Canadian Prime Minister Mark Carney flagged Anthropic's export control situation as a "model risk" concern in remarks to press this morning.
- IPO S-1 in flight: Anthropic filed its S-1 confidentially in early June. A prolonged regulatory standoff could complicate investor sentiment ahead of any public offering.
- Sovereign AI backlash: The blackout triggered calls across Europe and Canada for reduced dependency on US-controlled AI infrastructure. Former French PM Édouard Philippe: "Infrastructure controlled by others is infrastructure others can unplug."
Background: This is the first time the US government has used export controls to halt access to a commercial AI model already in wide public use. An administration official told Axios that future models crossing the same capability threshold would face the same pre-release review requirement.
Comments
Sign in to comment.