Trump posts AI video on X; debate about deepfakes

This week, Donald Trump shared several AI-edited posts on social media platforms X and Truth Social. A sensational video appeared authentic but was synthetically manipulated and lacked a clear label. It quickly spread to other platforms, reigniting the discussion about deepfakes in the political arena, particularly given the implications of the European AI regulation for governments and platforms.

The video contained artificially manipulated elements, making it extremely difficult for viewers to distinguish fact from fiction. This phenomenon is especially problematic in politics, where influential figures like Trump have a vast reach. The timing of such content, especially around political events, significantly increases the risks and impact. Deepfakes are a product of AI technology that uses generative models to create realistic fakes of images and sounds.

The European AI regulation (AI Act) requires compatible transparency for deepfake content. Creators and distributors must properly label synthetic media to prevent deception. Particularly for Dutch governments, labels must be added to all generated or edited media. Non-compliance can result in sanctions, and communication teams must therefore implement a proper labeling system and control mechanisms.

Furthermore, large platforms like X and YouTube must comply with the Digital Services Act (DSA), which obliges them to manage misinformation, such as that surrounding elections. However, enforcement is inconsistent, meaning that measures taken, such as labeling, are often implemented late or not at all. This creates an enforcement gap, especially when content is spread across other platforms. Initiatives like Community Notes on X and independent fact-checkers provide some support but are mostly reactive, which is problematic given the speed of generative content. In conclusion, experts emphasize that verification and transparent labeling are essential to ensure the integrity of imagery and maintain public trust.

Read the full article from AI Insider.