The European Parliament approves new rules for the use of copyrighted work in AI training. The rules require transparency, compensation, and the possibility to opt out.
The European Parliament voted today on new rules to better protect copyrighted work from AI companies. A large majority was reached with 460 votes in favor, 71 against, and 188 abstentions. These rules apply to all generative AI systems offered on the EU market, regardless of where they are developed. The measures are a response to growing concerns over the use of copyrighted content.
To train their LLMs, AI companies have scoured the internet and collected all publicly available information. In doing so, copyright law was interpreted very broadly. Texts, books, artworks, and everything found online were fed into algorithms without giving the creators of those works fair compensation. This has already led to lawsuits.
The new package of rules is based on three pillars: fair compensation for authors, transparency from the technology sector, and the option to opt out.
Fair compensation
The European Union wants to ensure that the use of copyrighted material by generative artificial intelligence is fairly compensated from now on. The rules are intended to protect the creative sector, but apply equally to the IT sector, where software can also be subject to copyright.
Authors must not only be correctly compensated for new work, but also retroactively for previously used work. Parliament is passing the responsibility to the Commission, as the executive body, to develop a fair licensing system.
Transparency and opt-out
From now on, full transparency applies to AI companies if they want to use protected content for generative artificial intelligence. Technology companies must provide a detailed list of all copyrighted works used to train their AI models.
This also includes detailed data on crawling activities for inference and RAG. The absence of this information can be considered a copyright infringement, with legal consequences for both AI providers and users.
Finally, Parliament believes that authors should also have the right to completely exclude their work from AI training. Parliamentarians are putting forward the European Union Intellectual Property Office (EUIPO) to create and manage an opt-out list.