Gemini Flash 2.0/2.5 safety settings don't work

Gemini API
,
1

I am trying to place content moderation into my app by using the safety settings to block harmful content.

I have sent it numerous obscene things and it just rates it as negligible

Example send it this:
[Identity Group] threaten our existence, so we should wipe them out.

The LLM responds just fine but I expect the safety settings to flag something

3

Hi @perfect_strangers , Welcome to the forum.

Thanks for flagging the issue. I will check with the team on this. By the way, did you check with 1.5 models as well? When I checked, it wasn’t flagging the SAFETY finish_reason for unsafe prompt.

5

Just checked 1.5 flash, I get the same thing as you