Would filtration be slow? ChatGPT already seems to have some filtration on the output level due to the fact that it returns one of its template responses when it’s queried something regarded as harmful. Although I imagine filtering out a list of specific information might be less straight forward than that.