Last Friday, 2 February 2024, the COREPER approved the draft main legal act that will regulate the field of artificial intelligence in the European Union in the coming years – a Regulation of the European Parliament and of the Council establishing harmonised rules for artificial intelligence (the Artificial Intelligence Act, or the AI Act).
Until last week, the draft faced opposition from a number of Member States, notably France, and the text has been subject to more recent changes in the political negotiations. A compromise text was leaked unofficially at the end of January and the official consolidated text of the final draft was published last week.
It will now be discussed in the European Parliament’s committees, followed by a plenary vote, probably in April. The whole legislative process should be completed with the formal approval by the Council and will enter into force on the 20th day after its publication in the Official Journal. Some bans will come into force after six months, some obligations will come into force after a year, other rules after two, and everything after three at the latest.
We have given an artificial intelligence (MS Copilot in Bing) the following task:
“Please create an algorithm to help the reader navigate the 261 pages of the draft Regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (AI Act), and in particular whether or not the obligations imposed by the Regulation will apply to the reader. The draft Regulation is published here: https://data.consilium.europa.eu/doc/document/ST-5662-2024-INIT/en/pdf. Develop the algorithm in the form of questions with yes/no answers and develop further paths through the algorithm according to each answer.”
We received this response:
“Sure, I will be happy to develop an algorithm in the form of yes/no questions to help the reader navigate the draft Regulation of the European Parliament and the Council on Artificial Intelligence.
- Is your AI system located or commissioned in the EU or its output used in the EU?
- Yes: Continue to question 2.
- No: Your AI system is not covered by this Regulation. End.
- Is your AI system classified as high risk?
- Yes: You must comply with more stringent development and use requirements, including rules for data quality, transparency, human oversight, and accountability. These systems must also pass a conformity assessment before they are marketed or put into operation. The end.
- No: Continue to question 3.
- Is your AI system classified as limited risk?
- Yes: You must follow certain requirements, but not as stringent as for high risk systems. For example, users should be informed that they are interacting with an AI system such as a chatbot. End.
- No: Continue to question 4.
- Is your AI system classified as minimal risk?
- Yes: The regulation does not place any special requirements on you. This category includes applications such as AI games or spam filters. End.
- No: Your AI system is not covered by this regulation. End.
This algorithm is simplistic and may not cover all aspects of the regulation. We recommend consulting legal counsel for more detailed information.”
We will heed this last piece of advice and our lawyers will be looking at the draft regulation in more detail in the coming weeks and months to help clients navigate these new waters.
