deep-analysis

Claude Fable 5 Returns in 19 Days: Why the First Commercial AI Switched Off by the U.S. Government Came Back on July 1 with New Safeguards

Anthropic's top-tier model, Claude Fable 5, returned globally on July 1, 19 days after its service was suspended. Launched on June 9, the service was completely halted three days later when the U.S. Department of Commerce imposed export controls following a 'jailbreak' report by Amazon researchers. Anthropic contested the basis of the action, arguing that other models yield the same results, and engaged in negotiations with the government for about two weeks. As a condition for the return, a new classifier capable of blocking the technique by over 99% was introduced, with Opus 4.8 responding in its place when blocked. Marking the first export control case applied to commercial AI, the incident leaves behind tasks such as refining AI regulatory procedures and the need for corporate model diversification.