First on the Scene: How the European Union is Leading the Charge to Regulate AI

The apprehension surrounding novel technology and the impact of it is not new; especially in cases where this introduction has had the tumult that Artificial Intelligence has endured. AI has served as an unnecessary boogeyman in the public consciousness. There is undoubtedly a potential unlike any other in this developing technology, yet we keep it at arm’s length. On the other hand, there is some warranted concern with such a powerful tool, but not from the tool itself. Like with any sea change in publicly available innovation, there will be those who abuse the gap between what is possible, and what has been legislated against.

Inevitably, bad actors will use the technology, and the law must stand in the way of malicious use. In the case of AI, this can be interpreted as even more sinister. At a time where it has become increasingly difficult to know reality from generated content, and knowing that it will only get harder, it makes building trust a battle. From mass misinformation, lifelike deepfakes, and the constant concern of how this will affect privacy, there is a distinct imperative for some legal guardrails.

This led the European Union’s (EU) Parliament to be the first to address this with a comprehensive answer to the issue of AI. On March 13, 2024, the Artificial Intelligence Act was adopted, with applications that pose challenges to developers no matter the country of origin, and hefty fines for rulebreakers. The EU’s objective with human-centered AI legislation is that users can enjoy the benefits, while retaining proactive safety features. With a focus on transparency, data quality, and oversight, the first step has been made with a broad definition of what AI is, and a risk mitigation system with varying levels of restriction.

The definition used in the legislation comes from the Organization for Economic Cooperation and Development (OECD) as (1) operating with various levels of autonomy, and (2) with the ability to make inferences from the input they receive to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. This broad description goes further than capturing just AI platforms and models. Any product, help bot, or recognition system that does this is brought in. The Act then describes the risk categories as minimal, limited, high, and unacceptable. Minimal is the lowest risk category and carries no restriction for the amount of threat it poses, like a spam filter. In contrast, with the unacceptable risk category, the use of that system is a clear threat to fundamental human rights. This covers any model that manipulates or exploits behavior with the goal of distorting it.

This legislation marks what is hopefully a purposeful change in global attitude when it comes to AI. Many more steps will need to be taken as this tool becomes better, but with this first, governments can use this example to build a crucial layer of protection.

Article Written by Johnnie Nicholas Conner IV

Sources:

Teri Schultz, The world’s first comprehensive artificial intelligence legislation is here, NPR Morning Edition, March 14, 2024, available at https://www.npr.org/2024/03/14/1238496174/the-world-s-first-comprehensive-artificial-intelligence-legislation-is-here (last visited Sep. 3, 2024).

Kirk J. Nahra, The European Union Adopts the AI Act, WilmerHale, March 14, 2024, available at https://www.wilmerhale.com/en/insights/blogs/wilmerhale-privacy-and-cybersecurity-law/20240314-the-european-parliament-adopts-the-ai-act#:~:text=On%20March%2013%2C%202024%2C%20the,transparency%2C%20human%20oversight%20and%20accountability. (last visited Sep. 3, 2024).

European Commission, European approach to artificial intelligence, European Commission (March 13, 2024) available at https://digital-strategy.ec.europa.eu/en/policies/european-approach-artificial-intelligence (last visited Sep. 3, 2024).

Gernot Fritz, EU AI Act unpacked #8: New rules on deepfakes, Lexology, June 26, 2024, available at https://www.lexology.com/library/detail.aspx?g=c25237ee-a37f-4959-837c-f32a624f54ab#:~:text=The%20AI%20Act%20does%20not,the%20users%20of%20AI%20systems. (last visited Sep. 3, 2024).

Artificial Intelligence Act, Parl. Eur. Doc. (COM (2021)0206) – C9-0146/2021 (2024).