The EU AI Act Newsletter #3
02/03/22-14/03/22
Welcome to the EU AI Act Newsletter, a brief biweekly newsletter by the Future of Life Institute providing you with up-to-date developments and analyses of the proposed EU artificial intelligence law.
Legislative Process
The Committee on Legal Affairs (JURI) at the European Parliament published their amendments on the AI Act. The committee recommends narrowing the definition of AI, providing high-level principles for European Standardisation Organisations, making the supply chain more balanced, especially with regard to general purpose AI systems, giving the AI Board a strong mandate and sufficient resources, and much more.
The Committee on Industry, Research and Energy (ITRE) at the European Parliament published their draft opinion on the AI Act. The committee recommends enhancing measures to support innovation; providing a concise and internationally recognised definition of AI; setting high but realistic standards for accuracy, robustness, cybersecurity and data; encouraging uptake of AI by emphasising social trust and value chain responsibility; and future-proofing the Act through better linkages to the green transition and future trends.
The IMCO/LIBE joint committee at the European Parliament announced a new timeline for the AI Act, with the deadline for amendments on 26 April 2022 and voting on the final amendments on 29 September 2022. However, a journalist at EURACTIV has since shared a new timeline indicating further delays:
The French Presidency of the Council of the EU has circulated a new compromise text on articles 53-55a of the AI Act, which focus on measures supporting innovation. Some of the changes are intended to clarify the conditions of AI regulatory sandboxes, provide rules for testing high-risk AI systems in real-world conditions, and make clear that support measures are for SMEs in particular, including start-ups. To our knowledge, the document is not yet publicly available.
Analyses
Orgalim, which represents Europe's technology industries, published their proposed amendments to the AI Act. They recommend simplifying the AI Act by focusing more on the protection of fundamental rights and exempting already regulated areas, while at the same time creating more flexibility for risk-based, case-specific implementation of the requirements. They urge against the over-regulation of AI so as to ensure Europe’s leadership in this area.
Wired published an op-ed advocating against the use of an overly narrow definition of artificial intelligence in the AI Act. The piece points out that some intelligent systems are at risk of being excluded from the legislation and argues that this is bad for both businesses and citizens. The original proposal uses a broad definition of AI but half of EU member states are pushing back against this definition. The op-ed claims that while the costs of regulation should be minimised, a very narrow definition will not ensure sufficient safety.
EDRi and 40 other members of civil society have called on the EU to prohibit the use of predictive and profiling AI systems in law enforcement and criminal justice in the AI Act. Their joint statement discusses the harmful impact of predictive, profiling and ‘risk’ assessment systems in law enforcement and criminal justice, and provides justifications for proposed amendments to the Act.