reneberlin
Joined 11 years ago
Karma
215
Submissions
286
Member since
May 9
Joined 11 years ago
I find it troubling that people with no prior knowledge begin "self-hosting" their chatbots like this and put their secrets and personal stuff there, which this repo encourages. From a security-view it's a nightmare to advertise this.
https://github.com/x1xhlol/system-prompts-and-models-of-ai-t...
https://xunroll.com/thread/2064776322979676227
Using combinations of jailbreaking-techniques including: writing cyrillic helped a lot to disarm the filter.
The typical user of something like this is probably not a developer or screenworker at all - which is fine. But then came our agents and lots of people try to use these pets for reflecting current agent-status. Maybe it's a taste-question, but for me, i can't seriously use any of these.