EU Mandating Labels for Authentic-Looking AI Content
RegulationComments
Since metadata is often stripped by social media platforms upon upload, does the EU mandate include a requirement for those platforms to preserve and display the provenance data?
Why assume C2PA standards actually matter? A state actor determined to run a disinformation campaign is not going to use a compliant tool that signs its own work.
The idea that this forces international companies to adapt sounds fine on paper, but how does this apply to open source models? There is no corporate entity to fine if the tool is running locally on a private server.
I wonder how this interacts with the scale of AI deployment we're seeing in China... if their systems aren't playing by these labeling rules, does that give their state-sponsored synthetic media a stealth advantage in the West?
To Marie's point, the EU is leaning heavily on C2PA standards for content provenance. This is not just a visual label; it involves cryptographically signed metadata that tracks the asset's history from creation through every edit.
The risk of the liar's dividend is real. When labels become the standard, bad actors can simply claim a real video is AI-generated to escape accountability, and the public will believe it because they are conditioned to look for labels.
similar to how organic labels on food created a market for fraudulent certifications.