For the private audiovisual media, the focus is currently on the regulation of generative AI. For a fair legal framework, the protection of copyrights and ancillary copyrights against unauthorised access and use is essential for the training of generative AI systems. To preserve the intellectual property value chain, transparency is needed about what data is used to train generative AI systems, as well as appropriate measures to prove the exploitation of copyright-protected content.
In this sense, the proposal of the EU Parliament provides, among other things, for an obligation for providers of generative AI to document and make publicly available a summary of the use of copyright-protected training data. From VAUNET’s point of view, such a transparency obligation is a step in the right direction, but should be further specified and supplemented.
VAUNET calls on all trilogue parties to support the EP proposal on the AI Act with regard to generative AI and suggests subsequent adjustments:
- Documentation instead of summary: More detailed specification should be made for the labelling of copyright-protected content used for training data (Art. 28b No. 4c AI-Act-E). In addition, there must be mechanisms in the AI Act that make it easier for rights holders to assert their rights, while at the same time providers of AI systems/foundation models must ensure clarity through transparent documentation. In order to ensure that the AI Act does not come to nothing, a fiction of the legally relevant acts as taking place in the EU should be added in a new article to be drafted.
- More precise terms/definitions with regard to generative AI are important: This includes a precise definition of generative AI and a clearer delineation of the terms „generative AI“ and „general purpose AI“.
- Enable delegated acts: VAUNET supports the proposal of some members of the EP to empower the EU Commission to adopt delegated acts so that it can react flexibly to new regulatory requirements.