Anthropic restores Fable 5 with added safeguards

Anthropic restored its Fable 5 AI model with enhanced safeguards after a two-week federal export-control freeze triggered by a jailbreak that identified software vulnerabilities. The Commerce Department imposed controls on June 12 following an Amazon researcher's report, but lifted them after Anthropic implemented a classifier blocking the exploit over 99% of the time. The incident prompted Anthropic, Amazon, Microsoft, and Google to jointly draft a severity framework for AI jailbreaks.

The real practitioner-relevant finding buried in this story is not that Fable 5 was uniquely dangerous, but that it wasn't: Anthropic's own post-incident testing found that weaker, generally available models, including Opus 4.8, GPT-5.5 and Kimi K2.7, could reproduce the same vulnerability-identification and even the same exploit-code demonstration that triggered a two-week federal export-control freeze. The Commerce Department applied export controls to Claude Fable 5 and Mythos 5 on June 12 after Amazon researchers reported a jailbreak that got Fable 5 to identify software vulnerabilities; because Anthropic could not verify user nationality in real time, it suspended both models for everyone. Mythos 5 returned for Project Glasswing partners on June 26, controls lifted June 30, and Fable 5 returns globally July 1 with a new classifier Anthropic says blocks the reported technique over 99% of the time, falling back to Opus 4.8 when triggered. More significant for the industry: Anthropic, Amazon, Microsoft and Google are jointly drafting a four-factor severity framework for AI jailbreaks, and Anthropic committed to giving the government pre-release model access going forward.