Negotiations in the Council of Europe on the Convention on Artificial Intelligence: the ongoing story of the mountain which labored and brought forth a mouse

Anonymous// When, in 2021, the Committee of Ministers of the Council of Europe (CoE) set up a Committee on Artificial Intelligence (CAI)1, tasked with the elaboration of a legal instrument on the development, design, and application of artificial intelligence (AI) systems, such an initiative received a strong support from all those who consider that the CoE’s standards on human rights, democracy and the rule of law should be the reference for an international binding instrument on AI, while allowing innovation.

But the initial enthusiasm was dampened quite quickly, first with the working methods of the CAI, which chose to establish a drafting group, composed only of potential Parties to the Convention, and therefore preventing de facto all observer organisations from witnessing on the spot the process of elaboration of the instrument. This exclusion was aggravated by the absence of publication of the first version of the Convention.

The lack of transparency of the working methods prompted civil society organizations to write to the CAI on 28 January 20232, in order to present their priorities. It is only at its 4th Plenary meeting, in February 2023, that the CAI decided to make the revised version of the Convention public3. Such a decision was anyway unavoidable, since the text had been leaked, but the Chair opened the meeting by deploring such breaches of confidentiality.

However, the fact remains that substantive discussions are taking place behind closed doors, even if civil society organisations and other stakeholders can provide their comments on the draft texts during the Plenary meetings. In the Plenary meeting of 31 May-2 June 2023, the participants were informed that informal meetings of the drafting group would be organised during the period from June to December 2023 in order to “explore possible solutions to some of the outstanding issues”, excluding once again observer organisations from the key discussions4.

Second, it appears that the ongoing negotiations are moving away from the initial objective, pursued by the Council of Europe, of elaborating a convention on AI, based on the Council of Europe’s standards on human rights, democracy and the rule of law. They are also departing from the outcomes of the work5 of the Council of Europe Ad hoc Committee on Artificial Intelligence (CAHAI), which was the predecessor of the CAI. In particular, the CAHAI considered that the future instrument should contain “certain fundamental principles of protection of human dignity and the respect of human rights, democracy, and the rule of law, which should apply to all development, design, and application of AI systems, irrespective of whether the actor is public or private”. The CAHAI also recommended that the instrument should provide for the establishment of a methodology for risk classification of AI systems, which would include “a number of categories (e.g. “low risk”, “high risk”, “unacceptable risk”), based on a risk assessment in relation to the enjoyment of human rights, the functioning of democracy and the observance of the rule of law”.

Unfortunately, the revised draft does not reflect such recommendations. The delineation of the scope of the future Convention is not totally clear as regards the inclusion on equal terms of the public sector and the private one: in particular, the choice to dedicate Chapter II of the draft to the application of AI systems “by public authorities” only creates some doubts as regards the final outcome of the negotiations on this crucial point. Moreover, there is no reference to the risk-based approach, nor to AI systems presenting unacceptable risks. At last, and this point is probably the most worrying at this stage, a number of articles refer to the domestic law of each Party, as if the application of the principles laid down in the future instrument should depend on pre-existing rules established at the national level. In such a case, how could the future Convention constitute the first legally binding framework for the regulation of AI, equally applicable to all Parties?

Such a shift is all the more surprising, given the recent Declaration of Heads of State and Government of the Council of Europe, in the Summit held in Reykjavík, Iceland, on 16-17 May 2023: they committed “to ensuring a leading role for the Council of Europe in developing standards in the digital era to safeguard human rights online and offline, including by finalising, as a priority, the Council of Europe’s Framework Convention on Artificial Intelligence”.

If the Council of Europe really wants to play a leading role in developing standards in the digital era, the future Convention on AI should create a legally binding framework based on the protection of human rights and ensuring legal certainty to all Parties. The wish to see observer States, including the USA and Canada, being part of the Convention cannot justify the reduction of the future instrument to a list of “principles” whose implementation may differ at the national level, depending on the domestic law. Otherwise, one could wonder what would be the added value of such an instrument, in comparison with the OECD AI Principles6, adopted in May 2019.

The DNA of the Council of Europe is the effective protection of human rights, and it would be wise for its high representatives to remember it.

2 Civil Society Statement on the Council of Europe Convention on AI, 28 January 2023:

3 CAI, 4th Meeting, list of decisions:

4 CAI, 6th Meeting, list of Decisions :

5 CAHAI, Possible elements of a legal framework on artificial intelligence, based on the Council of Europe’s standards on human rights, democracy and the rule of law, 3 December 2021:

6 OECD AI Principles: /en/ai-principles OECD AI Principles