The Once And Future Fable #2

Claims

• The White House forced Anthropic to take down Claude Fable and Mythos by imposing export controls after Anthropic refused to comply with a 90-minute undetailed takedown order.
• The jailbreak that triggered government action can be replicated on GPT-5.5 and other models without any bypass, providing no unique offensive capability that justifies restricting Fable specifically.
• Cybersecurity experts who reviewed the underlying research report concluded its findings were not a national security threat and were focused on defensive capabilities normal for any AI security tool.
• The government's action was partly motivated by political grievances including disapproval of Anthropic's political associations and its failure to perform sufficient deference to administration officials.
• The resulting de facto AI licensing regime is informal, opaque, and subject to ad-hoc enforcement by senior officials with no technical expertise, creating worse regulatory uncertainty than formal statute would.
• Project Glasswing and US intelligence agencies lost access to Mythos as collateral damage from the export control order, net harming American cyber defense capabilities.