cm0002Jul 8

Researchers Jailbreak AI by Flooding It With Bullshit Jargon

https://lemmy.cafe/post/20090640

Researchers Jailbreak AI by Flooding It With Bullshit Jargon - Lemmy Cafe

Lemmy

Show threadAvicennaJul 8
I wonder if they tried this on DeepSeek with Tiananmen square queries
Show threadSheeEttinJul 9
No, those filters are performed by a separate system on the output text after it’s been generated.
Show threadAvicenna
makes sense though I wonder if you can also tweak the initial prompt so that the output is also full of jargon so that output filter also misses the context
Jul 9 at 11:42amWeb
Show threadSheeEttinJul 9
Yes. I tried it, and it only filtered English and Chinese. If I told it to use Spanish, it didn’t get killed.