Author: Yasmin Morais, Foreign, Comparative & International Law Librarian – Collection Development at Harvard Law School Library
This post will discuss the European Artificial Intelligence Act (AI Act), and the responses of vendors to the Act.
With the objective of ensuring that there is responsible development and deployment of artificial intelligence, the European Artificial Intelligence Act (AI Act) entered into force on August 1, 2024. This follows the successful consensus in the European Parliament and Council in December 2023.
The AI Act focuses on the risks to the health, safety, and basic rights of EU citizens and outlines four risk categories:
- Minimal Risks – no obligations for spam filters, AI video games, etc. Companies can include codes of conduct.
- Specific Transparency Risks – Users must be informed of interaction with a machine, as well as the labeling of AI-generated content.
- High Risk – Strict requirements for AI-based medical software, and AI systems used for recruitment. Requires clear user information, human oversight, and risk mitigation systems.
- Unacceptable Risk – Threats from governments or companies to fundamental rights are banned.
The EU website has a particularly useful AI Act Explorer to facilitate searching of the Act. Further information is also available in the EU’s press release.
We asked three vendors that have implemented generative AI products if they have incorporated or assessed their compliance with the AI Act.
The following questions were posed to representatives of Thomson-Reuters, LexisNexis, and VLex:
- Is your company in compliance with or otherwise addressing Europe’s AI Act?
- Are there any provisions in the AI Act that may potentially impact your organization?
- What steps have been taken in response to the EU AI Act?
At the time of publication of this post, only VLex (see below) has provided a response. However, Thomson-Reuters and LexisNexis have promised to provide us with their responses later this week and we will update this post once we receive them.
VLex
“We can confirm that we are continuously following all recent legal developments in AI and, of course, the EU AI Regulation.
In line with the timeline set forth by the EU AI Regulation, we are in fact currently analyzing the impact of the AI regulation in Vincent. As you know, the measures that are potentially applicable to us will not enter into force until August 2025 or August 2026. We are however already working on identifying the measures that are applicable to us and we will soon have a calendar for their implementation. We are therefore on track to having all requirements implemented by the time they are into force and, actually, some of the requirements are already partially implemented.”
With respect to current standards, legislation, and guidelines within the United States, the following are noteworthy developments:
- American Bar Association Formal Opinion 512, identifies ethical issues and provides guidance for lawyers on generative artificial intelligence. (July 29, 2024.)
- California Governor’s veto of SB 1047 Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (2023-2024). (September 29, 2024). Among other provisions, the bill would have required developers to adopt a safety and security protocol, including the capacity to enact a full shutdown for safety reasons if needed. It would have also empowered the Attorney General to bring a civil action to enforce certain auditing requirements and protections for whistleblowers.
- Executive Order on the Safe, Secure and Trustworthy Development and Use of Artificial Intelligence (October 23, 2023) This Executive Order directs the Secretary of Commerce, in coordination with the Secretaries of Energy and Homeland Security, to establish standards and best practices for developing and deploying trustworthy AI systems. A plan drafted in accordance with this EO was published in July. National Institute of Standards and Technology (NIST) A Plan for Global Engagement on AI Standards (July 26, 2024)
Leave a Reply