Lawmakers in the European Parliament are one step closer to a way to combat generative AI. By the end of the year, the European Parliament hopes to reach a consensus on the draft law regulating AI.

The European Council already commented on the regulation in December. While most member states wanted to postpone the decision on how to deal with generative AI, the European Parliament wants to set strict guidelines.

Lobbyists and tech giants like Google and Microsoft have done everything they can to counteract these strict requirements in recent months. It is not yet known what the outcome of the negotiations will be, but Dragos Tudorache, the co-rapporteur on the EU AI law, hints in an interview with TechCrunch that the European Parliament is leaning towards a multi-layered approach.

The three levels of European AI legislation

The AI ​​law is expected to consist of three layers.

• A level to address responsibilities across the entire AI value chain
• A layer to ensure base models get some guardrails
• A layer to address specific content issues related to generative models like ChatGPT

On the current thinking of MEPs, the first of these three layers would apply to general AI (GPAIs) – large or small, fundamental or non-fundamental models – and aim to regulate the relationships in the AI ​​value chain.

The second proposed layer would concern fundamental models. Manufacturers of these base models would be subject to a number of specific obligations.

The third level proposed by the European Parliament focuses specifically on generative AIs, a subset of GPAIs consisting of language models, generative art and music AIs. According to the legislator, these instruments require even more specific responsibilities. For example, there should be requirements for the type of content these AIs can produce. The question of copyrighted material for training AI also continues to play a role.

The EU’s planned AI framework is likely to take some time, but the European Parliament is making progress. Parliament expects to reach agreement on a final AI law by the end of this year.