Or the AI companies could filter it from their training data. That would be another, probably easier, option.