@davidrevoy @EUCommission The platforms will most likely use watermarks in AI content. If calibrated correctly, they have very low false positive rates. However, they are also easy to remove and it is unclear what platforms will do with AI content generated with unwatermarked Open-Weight-Models etc.