“Since the rules were implemented, what we have seen is that while watermarks are there on AI-generated or AI-modified content, for the most part, these watermarks appear for only a fraction of a second in a video, or are positioned in a nearly invisible manner in photos that have been significantly modified with AI,” said a senior government official with direct knowledge of the matter, requesting anonymity since the proposal is still in the draft stages. The result of this is that the purpose of labelling AI content, which was to crack down on deepfakes and alert users about the same, is not being solved, the official added.
