US Blocks Fable 5 AI Access for Non-Citizens; Anthropic's Response

Summary

The United States government has ordered Anthropic to suspend access to its Fable 5 AI model for all non-US citizens, citing security concerns after Amazon researchers demonstrated a jailbreak that revealed minor vulnerabilities. Anthropic complied by pulling the plug entirely, making Fable 5 inaccessible to everyone. While Anthropic stated the discovered vulnerabilities were minor and discoverable by other public models, they also acknowledged the government's directive, though they expressed that the action did not align with their preferred transparent regulatory process. This shutdown comes as Anthropic has been a vocal advocate for government AI regulation, but the company also feels the government moves too slowly. Separately, Anthropic has rolled back a policy of "invisible refusals" for frontier AI model development, now making safeguard triggers visible to users. Users describe Fable 5 as a significant leap, shifting from prompt engineering to delegating complex tasks to a "thought and design partner" with judgment and taste, capable of methodical debugging and even generating functional software like "Concord" for data analysis, which researchers have desired for years but found unprofitable to develop.