(github.com)

536 points BlueFalconHD | 1 comments | 06 Jul 25 19:50 UTC | HN request time: 0.294s | source

I managed to reverse engineer the encryption (refered to as “Obfuscation” in the framework) responsible for managing the safety filters of Apple Intelligence models. I have extracted them into a repository. I encourage you to take a look around.

Show context

bombcar ◴[06 Jul 25 20:38 UTC] No.44483830[source]▶

>>44483485 (OP) #

There’s got to be a way to turn these lists of “naughty words” into shibboleths somehow.

replies(2): >>44484345 #>>44485000 #

1. spydum ◴[06 Jul 25 21:42 UTC] No.44484345[source]▶

>>44483830 #

Love idea, but I think there are simply too many models to make it practical?

↑

I extracted the safety filters from Apple Intelligence models