US Government Orders Shutdown of Anthropic's Powerful AI Models
Anthropic shut down its Fable 5 and Mythos 5 AI models on Friday evening after the Trump administration issued an export control directive citing national security concerns. The government cited a jailbreak vulnerability discovered in Fable 5 by Amazon researchers that could expose cybersecurity information. Anthropic received only 90 minutes notice and disagreed with the decision, arguing the vulnerability was narrow and already present in other public models like GPT-5.5. The shutdown disabled both models globally rather than just for foreign nationals, affecting hundreds of millions of users worldwide. This marks the first time the US government has used export controls to halt a commercial AI model already in public use. The incident reveals tensions between national security, AI innovation, and government regulatory authority.
Key facts
- Amazon CEO Andy Jassy alerted the White House Thursday night about a jailbreak discovered by Amazon researchers
- Anthropic received the government directive at 5:21 p.m. ET on Friday with 90 minutes to comply
- Fable 5 was released to the public on June 9, just three days before shutdown
- The export control order targeted foreign nationals but forced Anthropic to disable models for all 100+ million users globally
- Anthropic disputed characterization, saying the vulnerability was narrow and replicable using GPT-5.5
- This is the first time US government used export controls to halt a commercial AI model already in public use
- European leaders framed the shutdown as a wake-up call for digital sovereignty and technological independence
How different outlets framed it
Left-leaning coverage: Some analysts view this as justified government action to protect national security from dual-use AI technology that could enable cyberattacks. Others on the left emphasize that Anthropic's own safety warnings about the model should have made the company more cooperative, and that the Trump administration's earlier hostile rhetoric toward the company complicated negotiations around a reasonable safety fix.
Right-leaning coverage: Right-leaning framings through White House adviser David Sacks emphasize that Anthropic prioritized commercial release over safety, refusing a reasonable government request to fix the vulnerability. This narrative portrays the export control as a reluctant but necessary enforcement action, questioning whether Anthropic's public commitment to AI safety was genuine.
Limits of this briefing
The specific technical details of the jailbreak remain classified and not independently verified. The Trump administration has not released evidence supporting the severity of the vulnerability. Reports differ on whether Amazon tested models at government request or independently.
Sources
- Axios: How Amazon and the White House ended Anthropic's Fable Centre
- TechCrunch: Anthropic's safety warnings may have just backfired Centre
- The Conversation: Why the US government shut down Anthropic's latest Claude AI model Centre
- Fortune: A warning from Amazon led the White House to shut down Anthropic's Mythos model Centre
- NBC News: Anthropic suspends new AI models after government directive Centre