We introduce an initial framework for content traceability in AI-generated media, aligning with the objectives of the EU AI Act. The rapid advancements in generative AI (genAI) necessitate the development of reliable mechanisms for identifying and tracking AI-generated content to ensure transparency, trust and regulatory compliance. To address these challenges, we propose a conceptual infrastructure that facilitates media content registration for AI companies, artists and institutions. It enables provenance tracking and content authentication. Importantly, the proposed system is applicable not only to AI-generated content but also to non-AI-generated media. This dual functionality enhances trust beyond the requirements set forth in the EU AI Act by ensuring the identification of both authentic and synthetic content. The framework incorporates robust hashing techniques, digital signatures, and a database to mitigate the spread of media with uncertain provenance while adhering to regulatory guidelines. A key component of this approach is the adoption of the ISO-standardized International Standard Content Code (ISCC) as a robust hashing method. The ISCC’s decentralized architecture allows for independent implementation without legal constraints, and its adaptability ensures compatibility across various content formats. However, maintaining the flexibility to update hashing algorithms remains essential to address evolving technological advancements and adversarial manipulations.
Julian Heeger, Waldemar Berchtold, Simon Bugert, Martin Steinebach, "EU AI-Act: Tagging GenAI Content" in Electronic Imaging, 2025, pp 301-1 - 301-7, https://doi.org/10.2352/EI.2025.37.4.MWSF-301