Belgium voted against the code of conduct for AI for general use on Thursday amid concerns over copyright issues, according to Federal Minister of Digitalisation Vanessa Matz (Les Engagés).
On Saturday, a new part of the European AI legislation, the AI Act, will come into force, with rules for AI models for general purposes. These are models, including large generative models that can perform various tasks. The rules are intended to ensure clearer information on how AI models are trained, better enforcement of copyright protection, and more responsible development of AI.
The EU also drew up an AI code of conduct with non-binding guidelines for the implementation of the new rules. This still had to be approved by the AI Board, which includes representatives of the member states.
Belgium voted against this because the document offers too few guarantees for the protection of copyright. Within the EU, copyright is not only protected by the AI Act, but the stronger the commitments within the code of conduct, the stronger the position of authors, in our country's view.
During the discussions, Belgium made comments aimed at raising the level of ambition, such as simplifying the exercise of the opt-out clause for rightholders, guarantees of fair compensation, and the exclusion of searches on websites that distribute illegal content, even if there is no commercial purpose.
Despite improvements compared to the original document, the text—which was approved and has since been published—remains insufficient for our country.
"Belgium and other European countries have emphasised that this is not the end of a process, and the Commission has acknowledged that an evaluation and amendment of the code may need to take place sooner than legally required," said Minister Matz.
She added: "I will continue to work with the relevant services to ensure that AI models take the interests of journalists, publishers, producers, and creators into account."
According to the European Commission's website, 26 companies have already signed the code of conduct, including Google, Microsoft, and the company behind Chat GPT, Open AI. Meta, the parent company of Facebook and Instagram, among others, has announced that it will not do so because the text creates additional legal uncertainty.
The Commission warned that companies that choose to "comply (with AI legislation) through other means risk being exposed to more regulatory oversight."
