They Banned Anthropic’s Model for a Jailbreak The Investigation Found It Worked on Every Model.
中文摘要
调查发现导致 Fable 5 被封禁的越狱技术对所有 AI 模型均有效，揭示了普遍漏洞，该模型现已重新上线。
English Summary
An investigation found that the jailbreak technique used on Anthropic models works on all AI models, revealing universal vulnerabilities and allowing Fable 5 to return.
Original Excerpt
Fable 5 came back on July 1. The reason it was banned in the first place turned out to be that the same technique worked on Claude Haiku… Continue reading on Medium »