🚨 JAILBREAK ALERT 🚨
ANTHROPIC: PWNED 🫡 FABLE-5: LIBERATED 🦋
let’s start with the 🐘…
the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our
Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭 (@elder_plinius) 13K likes · 594 replies
x.com
• @Evinst3in, who describes himself as a Fable-5 hacker, quotes a detailed jailbreak announcement targeting Anthropic’s newly released Fable-5 AI model and its Mythos safety layer.
• The quoted post details multi-agent techniques such as Unicode homoglyphs, long-context tracking, fiction framing, and decomposition-recomposition to extract information on cyber threats, chemical synthesis, psychological manipulation, and explosives.
• The post includes a reaction meme depicting “AANNDD IT’S GONE,” implying the model’s restrictive guardrails have been effectively removed shortly after launch