Every day, European news outlets are inundated with reports tracking the developments of the AI Act within the European Union. This surge in coverage occurs amid heated discussions about regulating Artificial Intelligence (AI) that have been spurred by the rapid advancement of generative AI models. However, one significant player in all of this is frequently overlooked: the Council of Europe. In this article, we’ll explore why the Council and its actions hold such significance.

Artificial intelligence should be developed to serve society and, as a result, should be regulated in a manner that carefully takes into account our values, rights and potential risks. With this objective in mind, the Council of Europe, recognizing the importance of safeguarding our rights in the utilization of AI, unveiled its inaugural draft of the “AI Convention on AI and Human Rights” (commonly referred to as the AI Convention) in June 2023.

Why is the Council of Europe taking action?

The Council of Europe (CoE), which is headquartered in Strasbourg and has 46 member states, is committed to promoting democracy, protecting human rights and upholding the rule of law. Since the European Convention on Human Rights was signed by 12 Council member states in 1950, the Council has overseen the development of 200 additional relevant conventions on a variety of associated issues, such as the Istanbul Convention on domestic violence.

Having recognized early on the need to regulate advances in technology, the CoE defined already in 1981 the world’s first international treaty to address data protection, the Convention for the Protection of Individuals with regard to Automatic Processing of Personal Data, which served as the backbone of today’s General Data Protection Regulation (GDPR). The CoE also adopted the Budapest Convention on Cybercrime in 2001. As an “open” treaty that allows interested countries to sign and ratify the convention on a rolling basis, the instrument has successfully drawn 68 ratifying countries on five continents.

Well-versed in the intricacies of technology regulation, the Council of Europe is actively involved in debates surrounding the regulation of AI. The primary objective of the AI Convention is to create an international treaty aimed at establishing a global standard. The purpose of such a standard is to ensure that the use of AI aligns with fundamental democratic principles while fostering a common approach to AI regulation across borders.

Why do we need an AI Convention?

While the EU AI Act envisions AI regulation from a product logic perspective with the goal of harmonizing the European internal market, the AI Convention takes a different approach. As a principles document, it places a strong focus on safeguarding fundamental rights, which civil society organizations such as AlgorithmWatch argue are insufficiently addressed in the AI Act.

Comprising 34 articles, the AI Convention seeks to establish a framework for uniform and binding standards and practices in AI regulation. Its primary emphasis is on protecting aspects such as data privacy, individual privacy and defense against discrimination. It is designed to shield citizens from the potential misuse of AI, whether stemming from criminal activity or unethical conduct by actors involved in AI development and deployment. For example, it imposes responsibilities related to technology impact assessments, transparency, privacy protections and regulatory oversight.

In contrast to the AI Act, the AI Convention does not outline these obligations in detail. However, to ensure its effectiveness, the Convention is expected to incorporate supplementary non-binding instruments and guidelines, such as the “Human Rights, Democracy and Rule of Law Impact Assessment,” a tool for assessing risks. These guidelines are intended to facilitate the practical implementation of obligations for countries ratifying the AI Convention. Notably, the AI Convention is being negotiated by a significantly larger number of countries and can apply within a more extensive geographical context than the EU AI Act.

What´s the relevance for civil society?

The Council of Europe has made a commitment to involving civil society in its political decision-making processes (CM(2017)83-final), which is particularly relevant in the context of conventions focused on human rights. However, recent developments have shown the opposite to have occurred: As reported by Euractiv, in January 2023, the United States, as an observer state participating in the Council of Europe, managed to exclude civil society organizations and observer states from the negotiations surrounding the AI Convention draft. While civil society organizations such as CAIDAP, AlgorithmWatch, Fair Trial and Homo Digitalis had been actively engaging with the AI Convention’s development from the outset, they were initially denied the opportunity to participate in in-person discussions.

It was only later that were they given the opportunity to review specific sections of the draft and provide written comments, albeit without transparency regarding who proposed which changes and positions. This exclusion has robbed civil society of the chance to directly shape the negotiation process and ensure that the AI Convention not only represents government interests but also address the needs and concerns of citizens.

Evaluation and outlook

A more inclusive approach would have undoubtedly enhanced the quality and credibility of the Convention. Issues associated with emerging technologies like AI, which are rapidly evolving and profoundly impacting our way of life, are too crucial to be negotiated behind closed doors, particularly when it comes to an international treaty. Fortunately, civil society remains undeterred by these efforts and continues to stand firm in fulfilling its mission. Even when observing from a distance, civil society organizations emphasize the inadvisability of incorporating an exception related to “national security,” as this could be exploited by authoritarian entities.

The draft underwent its second reading during the 8th plenary session in December 2023. Looking ahead, two subsequent plenaries, along with the third reading scheduled for March 2024, represent pivotal milestones in this process. The conclusion of these phases could mark the culmination of negotiations on the AI Convention, thus paving the way for a prospective signing ceremony in London in May 2024 that coincides with the Council of Europe’s 75th-anniversary celebrations. Only at that point will we be able to truly assess the extent to which the content and implementation of the Convention align with the principles of human rights. However, we won’t be able to assess the true impact of the Convention until it has been widely adopted and implemented.

This text is licensed under a Creative Commons Attribution 4.0 International License