op: are you using various models in the AI responses? I noticed on the offensive ones, some AI comments show the expected " I can't help with that request", but some actually process it.
Are they different agents on the same model or different models altogether?
Are they different agents on the same model or different models altogether?