California’s AI Transparency Act requires major generative AI providers to embed hidden, traceable provenance information in AI-made images, videos and audio. The measure aims to make suspected deepfakes easier to investigate by showing which system created a file and when, rather than relying solely on whether the content appears convincing.

The law took effect on 2 August and introduces further obligations for large online platforms from 2027, followed by new requirements for newly produced phones, cameras and voice recorders in 2028. These digital fingerprints can provide a record of a file’s origin and history, but they cannot establish whether the claims or message within the content are true.