[co-author: Ilaria Boschi]
As we saw in our last bite, artificial intelligence has been (and is expected to remain) a hot topic this year. This article aims to outline the most relevant provisions envisioned by the EU institutions after the milestone of the White Paper on Artificial Intelligence, adopted by the European Commission in February 2020 (see our previous bite on this). As we will argue below, such provisions, including especially the so-called “October Framework on AI” of October 20, 2020, lay the foundations for a broader, harmonized, horizontal regulatory framework on AI, which is expected to be published soon.
So, while we are all waiting for the new draft EU regulation on AI, here is a list of references, our “AI starting package”, that will give us additional background to better understand the forthcoming regulations.
The Assessment List for Trustworthy Artificial Intelligence (July 17, 2020), also known as ALTAI, is the outcome of a two-year study conducted by the High-Level Expert Group on Artificial Intelligence on behalf of the European Commission. Its purpose is to guide experts and researchers of future AI-driven solutions through the implementation of the principles expressed in the Ethics Guidelines of April 2019, i.e.: i) human oversight; ii) technical robustness; iii) privacy and data governance; iv) transparency; v) diversity; vi) societal and environmental well-being; vii) accountability. In so doing, the “assessment list” also provides lawmakers and practitioners with the priority points that the European legislation is expected to regulate.
Indeed, the first parliamentary resolution that opened the above mentioned “October Framework on AI” covered “ethical aspects of artificial intelligence, robotics and related technologies” (October 20, 2020), and to a great extent followed the very same path outlined by the AI High-Level Expert Group. Furthermore, the regulation proposal also sets fundamental principles for good governance of AI systems and AI-driven technologies. Among others, such principles include:
With its regulation proposal on civil liability (October 20, 2020), the European Parliament called on the Commission to propose a regulation laying down rules for the civil liability claims of natural and legal persons against operators of AI-systems. The requested regulation proposal will apply on the territory of the European Union to all types of AI system operations, regardless of the location of the operation and of its physical or virtual nature. Among the most important provisions, there are several special rules addressed to operators of so-called “high-risk AI systems”:
It should also be noted that the operator of an AI-system that does not constitute a high-risk system should be subject to a standard “fault-based” liability as provided under the historical Product Liability Directive.
The last parliamentary resolution completing the October Framework provides guidance on IPR developments within the AI sector. The resolution calls for a two-fold approach: upstream, European patent law must be reviewed / re-examined so as to allow for the registration and protection of AI-based technical innovations; downstream, the EU must be ready to face a historic shift in the theory of creativity, extending its protection to “artificial creations” and thus granting copyright to the natural person(s) (developers or further processors of the AI system) that legally “publish” those creations.
The draft resolution published on January 20, 2021, aims at solving strategic questions of interpretation and application of international law within the EU in the areas of military and non-military uses of AI and robotics. The European Parliament specifically stressed the necessity of human oversight of any military use of AI systems and reiterated its call to ban AI-enabled autonomous lethal weapons (so-called LAWS). Furthermore, the provisional resolution expands on the AI implications over a number of civil and public uses, for which it demands that more regulations be adopted; that includes uses in the field of justice and health, in transportation and autonomous driving, and in security and mass surveillance.
As the latest addition to the AI regulatory framework, on March 16, 2021, the European Parliament announced that it is currently working on a new resolution governing AI technologies in the field of education, culture and audiovisuals (see here the ‘draft motion for a resolution on AI in the field of education, culture and audiovisuals’). Although not much information has been officially disclosed, the final vote on the resolution should be held this May, and the text will probably address the issue of reducing “gender, social and cultural bias” in AI technologies, particularly with regards to audiovisual streaming services. More specifically, the resolution proposal will likely address measures to fight disinformation and fake news, while strengthening inclusion and protection of minors, consistent with the recent debates on media, entertainment and e-sports (see more bites on this topics in our 19th TMT Bites review).
One year after the White Paper, all the above has provided us with significant background for the forthcoming draft General Regulation and Coordinated Plan on AI, which, according to a number of commentators, is expected in the second half of this month. This follows public consultations carried out last year which involved more than 1,200 contributions from EU citizens, member states, civil society, industry and other stakeholders. Given these figures, it is not surprising to see many eyes awaiting yet another milestone in the European regulatory path towards AI “excellence and trust”.