@dansup what are your thoughts on some on-device models that can help detect harmful content? Thinking of creating a Swift Package to handle a suite of these cases, if you'd like, for the iOS variant.