Anthropic Models Back Online After 20-Day Trump Clash
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic's models Mythos and Fable returned online July 1 after Amazon — its partner and investor — flagged a "jailbreaking" technical flaw that triggered sweeping Trump administration export controls and forced the models offline for roughly three weeks.
- Commerce Secretary Howard Lutnick called Anthropic CEO Dario Amodei on June 12 at President Trump's direction to demand a fast resolution, and when Amodei called back after receiving the export-control letter, Lutnick confirmed the goal was indeed to take the models offline.
- The federal Center for AI Standards and Innovation and the NSA rejected Anthropic's initial fixes, prompting further iterations before various agency heads approved the changes and cleared the July 1 release.
- Treasury Secretary Scott Bessent was the first administration official to learn of the jailbreaking concerns, worked with White House Chief of Staff Susie Wiles to re-engage the company, and helped push a cybersecurity executive order to completion.
- Cybersecurity experts later wrote an open letter to the administration stating that other leading AI models share the same jailbreaking vulnerability Amazon flagged in Anthropic's systems — undercutting the case for singling the company out.
- OpenAI's GPT-5.6 remains on hold and had no visibility into the Anthropic-White House discussions, with OpenAI now engaged in separate daily technical talks about its own model release.
Why it matters: The standoff demonstrates that frontier AI model releases can now be stalled by multi-agency export-control reviews triggered by a single partner's flag — and with OpenAI's GPT-5.6 also on hold, the timeline for getting models to allies, which proponents tied to competing with China, remains unresolved.
