Provenance marking now has to survive generated audio too
Article 50 requires synthetic audio, image, video and text to be machine-readable and detectable as generated. Models now producing synchronised stereo alongside frames make that a four-channel problem, and providers already on the market have until 2 December to comply with marking specifically.
Watermarking discussions have overwhelmingly concerned images, where the techniques are mature and the attacks well catalogued. Audio provenance is a less developed field with different physics — resampling, compression and re-recording degrade signal differently than cropping and re-encoding degrade pixels.
Joint generation raises a question nobody has a settled answer to: whether a clip carries one mark or two. Mark them separately and an attacker strips one channel and keeps the other. Mark them jointly and the mark has to survive both pipelines, which are routinely processed by different tools at different stages.
The December deadline for existing products is the regulator conceding that retrofitting provenance is a different order of work from adding a disclosure banner. It is a realistic concession, and it also means most synthetic media in circulation this year carries no machine-readable mark at all.
European Commission — Code of Practice on Transparency of AI-generated Content → · EU Artificial Intelligence Act — The EU AI Act's transparency rules: a practical guide to Article 50 → · European Commission — Commission starts enforcing AI Act rules and new transparency requirements on 2 August →