Trump Administration Orders Anthropic to Block New Models, Sparking Policy Debate
AI 안전 | Mon Jun 22 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 1 sources
An export control directive forced Anthropic to take two of its latest models offline, escalating debates over AI policy and digital sovereignty.
Analysis
[Anthropic] took Fable 5 and Mythos 5 models offline [1]
- Complied with the Trump administration's export control directive
- Decided on full shutdown due to inability to selectively block foreign users
- Fable 5 was for general public release
- Mythos 5 for existing users
[Trump Administration] issued an export control order based on national security concerns [1]
- Took swift action on Friday afternoon
- Coincided with the timing of Iran war negotiations
- Specific security concerns kept confidential
[Amazon] discovered Fable 5 guardrail bypass and reported it to the White House [1]
- Amazon researchers discovered methods to bypass guardrails
- CEO Andy Jassy raised concerns directly to the White House
- Identified as the direct trigger for this action
[Cybersecurity Expert Group] released an open letter urging Trump to rescind the order [1]
- Signed by leading cybersecurity experts
- Concerns over stripping advanced security capabilities from network defenders
- Warned the action is actually dangerous
[AI Industry Competitive Landscape] highlighted possibility of Anthropic being singularly targeted [1]
- Background of Anthropic's deteriorating relationship with the Trump administration
- Other major AI labs unlikely to face similar sanctions
- Possibility of paradoxical PR benefits also raised