Reposted by Ypsi!
Full unredacted methodology, now that Google's patched* the underlying bug
recursion.wtf/posts/jitor_...
*The specific jailbreak here is no longer live, but I have some concerns about the larger vulnerability class
recursion.wtf
Gemini JiTOR Jailbreak: Unredacted Methodology
How I taught Gemini to build its own euphemisms on the fly, bypass its safety filters, and comply with prompts it would otherwise refuse. Full methodology, now that the specific payload is patched.
![A drawing by @raymoldy. An anthro creature is holding up a laptop and excitedly telling the viewer "check it out, I've simulated my own consciousness!"
On the laptop screen is an image of that creature screaming, accompanied by a popup box that says "Screaming [OK]"](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:iln2wm3y343rxqhqbidgwse3/bafkreifq55enkyqi4piflfgxvmem63gbmxj4zzamg4p3nymcrqykxtwmjm)
