←back to thread

I extracted the safety filters from Apple Intelligence models

(github.com)
536 points BlueFalconHD | 3 comments | 06 Jul 25 19:50 UTC | HN request time: 0.593s | 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.
1. Ey7NFZ3P0nzAe ◴[07 Jul 25 07:22 UTC] No.44487579[source]▶
Well it's one thing to regex filter "boris johnson" but i see that "chatgpt" is filtered too and that's f*** up:

https://github.com/BlueFalconHD/apple_generative_model_safet...

replies(2): >>44489707 #>>44494268 #
2. Ey7NFZ3P0nzAe ◴[07 Jul 25 12:32 UTC] No.44489707[source]▶
Ffs it's also rejecting french words related to being poor or immigrant or even welfare:

https://github.com/BlueFalconHD/apple_generative_model_safet...

Aide sociale Chomeur Sans abri Démuni

That's insane!

3. kridsdale1 ◴[07 Jul 25 20:19 UTC] No.44494268[source]▶
“Gemini” is in there too.