On EU AI Office

Assimakis Komninos, a partner at White and Case LLP, writes about the establishment of EU AI Office on his LinkedIn post. He writes:

One particular development that has gone relatively unnoticed is the publication last week by the European Commission of its Decision establishing the European Artificial Intelligence Office (AI Office). So what is this?
In December, the Council of the European Union presidency and the European Parliament’s negotiators reached a provisional agreement on the text of the AI Act, which has been published (or rather: leaked) in various sites. The big change from the Commission’s Proposal is the provisions on general-purpose AI models and the introduction of a Commission competence to enforce these rules (through?) an “AI Office”. One more area of enforcement for the Commission!
I checked the text of the trilogue version of the AI Act and last week’s Commission Decision. I was perplexed.
First, the Commission Decision establishing the AI Office does not sound like very much enforcement-related to me. It looks more like a task force internal to DG CONNECT that will be doing advocacy work.
Second, the current trilogue text of the AI Act is very perplexing in that regard. Articles 52a and 52b echo Article 3 of the Digital Markets Act (DMA) and provide for a system of designation. Then, you have specific obligations and again an experienced eye will see the parallelism in the structure of these obligations with the DMA (of course the content is totally different). But where do we find the powers that the Commission (?) or the AI Office (?) will have? This is where the real mess starts. Article 63a, which gives the impression of having been drafted at the last minute, provides: “Where an AI system is based on a general purpose AI model and the model and the system are developed by the same provider, the AI office shall have powers to monitor and supervise compliance of this AI system with the obligations of this Regulation. To carry monitoring and supervision tasks the AI Office shall have all the powers of a market surveillance authority within the meaning of the Regulation 2019/1020”. But then, there is a new Chapter entitled “Supervision, investigation, enforcement and monitoring in respect of providers of general purpose AI Models”, which adds to the overall confusion. Sometimes it mentions “the Commission”, other times it mentions “the AI Office”. There is a patchwork of provisions on investigatory measures, even on commitment decisions; I have never seen such unsystematic drafting. Finally, you have Article 72a, which is about imposing fines going up to 3% of the total worldwide turnover of the provider of general purpose AI models…
Honestly, I would just pause the process and rework entirely this chapter. It’s a whole mess. Surely a look at the relevant chapters in the DSA and the DMA would help the drafters…
Perhaps the Commission knew as much when it said at the end of Recital 9 of its Decision that “once the Regulation is adopted, this decision may be revised”…

Please click on this link to read the full post.

Your account