Artificial Intelligence: will the Union be able to exploit its benefits?

By Domenico Sorrentino, Research Assistant, Public Policy Brussels

Artificial Intelligence is developing by leaps and bounds. It already promises to change our lives in a dizzying number of ways: by improving healthcare, increasing the efficiency of farming and production systems, contributing to climate change mitigation, and increasing European citizens’ security, among other possibilities. At the same time, however, advancing AI also entails a number of potential risks, such as increasing intrusion into our private lives, gender-based or other types of discrimination, and feeding into opaque decision-making.

The Czech Presidency of the Council of the EU has made the Artificial Intelligence Act one of its priorities in the digital sector. The coming months will tell whether Prague is able to reconcile the different positions of EU Member States.

The EU is acting to address the opportunities and challenges of AI from the perspective of what it describes as “European values,” including the protection of individual privacy.  In April 2021, the Commission presented a proposal for a Regulation that would apply broadly to AI, and which sets out a series of definitions. Some of these, such as the AI system definition, remain the subject of intense debate among MEPs considering the proposal, reflecting the complex nature and scope of the issues at hand.

Broadly speaking, through this Regulation the Commission intends to establish a legislative framework which would harmonise rules for AI systems in the EU as well as establish certain requirements and prohibitions. Those prohibitions would apply to AI systems which have a significant potential to manipulate people, by exploiting vulnerabilities of specific vulnerable groups in order to materially distort their behaviour. The proposal also prohibits AI-based social scoring for general purposes done by public authorities. Thirdly, the use of ‘real-time’ remote biometric identification systems in publicly accessible places is also prohibited.

This is by no means the only subject of debate among legislators, however. Another is the establishment of a European Artificial Intelligence Board and the need for centralised governance and harmonisation across Member States with the aim of enabling companies to operate more smoothly. Others focus on the divergent approaches taken by MEPs, with some seeing the proposal through the lens of innovation, and others seeing it as a human rights initiative, with an emphasis on establishing new compliance mechanisms.

What are the human rights implications of artificial intelligence?

MEP Dragoș Tudorache, Rapporteur on the Artificial Intelligence Act for the Committee on Civil Liberties, Justice and Home Affairs, outlined the importance of having an innovation-friendly regulatory environment for AI. He noted “transatlantic cooperation in setting international standards within multilateral fora is essential for our future”.

Industry, for its part, has warned against an overly restrictive approach to AI. These models are not inherently high-risk, as they do not yet always have an intended use and can be used in a wide range of applications, including low- or no-risk uses. There are also knock-on effects to be considered across different fields, even beyond AI.

Conversely, Daria Honitiu, Research Associate at the University of Edinburgh, argues that adopting a human rights perspective would allow the proposed framework to better protect the safety, autonomy, and dignity of citizens. According to her, we need to bring human rights back into the equation of how emerging technologies can shape our understanding of personal autonomy and human dignity.

Intellectual property rights in AI technologies

Human rights are not the only source of concerns. When we examine outputs or creations developed with the support of an AI or based on AI technology, we are moving into the field of software, where protections necessarily intertwine with the world of intellectual property (IP). It is therefore important to understand what possible systems of IP protection apply to artificial intelligence. Bearing this in mind, the European Patent Office has updated its guidelines to include a specific section on the patentability of inventions involving AI, also defined as “computer-implemented inventions” (CII).

The legislative discussion is followed by the first steps in the field of research and testing. A clear example is the development of “sandboxes”, with an already up-and-running Spanish government’s Regulatory Sandbox on Artificial Intelligence, recently launched with the European Commission support. This sandbox will allow businesses to explore and experiment with new products and services under a regulator’s supervision. It provides innovators with incentives to test their innovations in a controlled environment and allows regulators to better understand the technology itself. The amendments proposed by the Rapporteurs of the AI Act go in the direction of strengthening sandboxes’ role and their presence in each Member State, at all levels – local, regional and European.

Even as the AI Act progresses, another EU legislative initiative, a revision of the Product Liability Directive, could also raise questions related to AI. MEP Svenja Hahn has flagged the file in question as a source of potential challenges, in particular the question of ensuring coherence for AI liability and the need for a central oversight authority across two pieces of legislation with independent processes.

Please reach out to us at brussels@instinctif.com if you are interested in receiving updates on the changes taking place or need help weighing in on these initiatives.