AIHacker Noonabout 9 hours ago
Fable 5 Jailbreak Shows AI Safety Limits
3 min read
Anthropic's Claude Fable 5 was jailbroken again within days of returning online. Security researcher Vitto Rivabella bypassed parts of its safety system after 20 hours of testing. The model blocks about 90% of attacks but remains vulnerable at the edges, especially in less common languages. Anthropic opened a HackerOne program called Cyber Jailbreak for reporting cyber misuse vulnerabilities.
Level
Hype check
Tap to vote and see what everyone thinks.
