Yesterday two EU committees, the Internal Market Committee and the Civil Liberties Committee, adopted the negotiating mandate for the EU AI Act, including adding some quite impressive amendments[1]. This might be the start of the ultimate phase of the EU Artificial Intelligence Act.
The EU’s first try and regulate Artificial Intelligence has come a great distance: When the EU Commission introduced its first draft of the act in April 2021 many — including me — expected the act to be finalized by summer 2022, but since then many details of the act have been fiercely debated.
The committees also agreed on an inventory of amendments which make clear and extend the AI Act, in some features actually quite dramatically [2]. Some highlights:
- The list of of intrusive and discriminatory uses of AI systems banned by the act now includes distant biometric identification systems in publicly accessible spaces, biometric categorisation systems (e.g. categorizing by gender, race, ethnicity, citizenship status, religion, political orientation) and using AI for predictive policing.
- AI systems which might influence voters in political campaigns and by use of suggestion systems on very large platforms (as listed within the Digital Service Act[3]) have now been added to the “high-risk” category.
- Recent transparency and risk assessment requirements for providers of (generative) foundation models like GTP.
- Clarified exemptions for research.
Especially interesting are the brand new requirements for using AI in suggestion systems and the specific mention of foundation models.
It seems obvious that a few of the amendments are a response to the developments during the last 12 months. And while plenty of hype has been created within the media around ChatGPT & Co, it has indeed grow to be increasingly clear what the true life impact of huge language models might appear like.
Some have criticized the AI Act for branching out into areas already covered by the GDPR and overlapping in parts with the Digital Service Act (e.g. on algorithmic transparency), but such overlaps seem unavoidable for a regulation covering such a various field of applications.
It’s encouraging that transparency requirements for AI generated content seems not limited to simply images any more, and the increased transparency requirements usually will serve the aim a lot better than regulating very specific use cases.
Moving the usage of AI by very large platforms into the high risk category is smart and acknowledges the impact these platforms can have. One can only speculate if the takeover of Twitter has helped sharpen the EU’s mind on the matter. Astonishingly, there was plenty of focus within the EU almost about suggestion algorithms [4] recently, something which was quite missing from the early drafts of the AI Act.
While the AI Act’s references to copyright issues in generative AI are still very vague and only stress how much of a gray area it’s, requiring providers of huge models to be more transparent about their sources seems not a nasty thing as such. As many features of the act it is going to be seen how this works out in practice.
So is that this it, is that this the homestretch now? Possibly. Before negotiations with the Council on the ultimate type of the law can begin, the draft negotiating mandate must be endorsed by the European Parliament, with a vote possibly happening in the course of the 12–15 June session. And while it’s possible that some groups in parliament will attempt to water down or amend the act again, the clear cut voting ends in the committees (84 votes in favour, 7 against, 12 abstentions) seem to be a broad consensus has finally been reached. So we could see the act finally coming this 12 months. Probably.
[1] Compromise Text (consolidated version 11.05.2023)
[2] A more detailed list of amendments
[3] https://ec.europa.eu/commission/presscorner/detail/en/IP_23_2413
jazz work coffee
TornadoCash: Your gateway to anonymous transactions on Ethereum. Keep your financial activities private with this decentralized mixer