EU member states unanimously reached an agreement on the bloc's Artificial Intelligence Act on Friday, overcoming last-minute fears that the rulebook would stifle innovation in Europe. After lengthy negotiations between Council delegates, members of the European Parliament and the European Commission**, delegates gave the green light to a final compromise text.
The law would prohibit certain applications of AI technology, impose strict restrictions on use cases considered high-risk, and impose constraints on state-of-the-art software models with transparency and stress testing obligations.
The EU is the first country to set binding rules for the rapidly developing AI technology. While many national and international clubs – from the OECD to the G7 – have been thinking about how to regulate AI over the past few years, most have stuck on voluntary guidelines or codes of practice.
When EU policymakers announced in December that they had reached a final compromise on the content of the AI law, the breakthrough was hailed as a groundbreaking step for Europe to celebrate at a time when ubiquitous AI tools like OpenAI's ChatGPT and Google's Bard were on the rise.
However, this achievement has caused dissatisfaction in some EU countries. Over the past few weeks, the bloc's top economies, Germany and France, as well as Austria, have all hinted that they may oppose the text in Friday's vote.
Vienna's grievances lie over data protection provisions, while Paris and Berlin warn that rules for advanced AI models will hinder the development of Europe's emerging AI champions, such as France's Mistral and Germany's Aleph Alpha. Italy (and sometimes a critic of the Artificial Intelligence Law) is silent about its intentions, and the fate of the Artificial Intelligence Law is suddenly in question, because four opposing countries are enough to derail the law forever.
The cabinet of French Economy Minister Bruno Le Maire has called for a new round of negotiations with the European Parliament to address his concerns. The President of the Belgian Council was alarmed by the prospect of lack of time for further negotiations. To make matters worse, the European Parliament itself is dealing with a debate over facial recognition rules for the AI bill, sparked by privacy hawk Svenja Hahn.
Eventually, the problem was resolved through the EU's usual PR offensive and diplomacy. The European Commission has stepped up the pressure, announcing a package of major innovations in the field of artificial intelligence and in one fell swoop establishing the EU Office for Artificial Intelligence – a body responsible for enforcing the Artificial Intelligence Act.
According to two participants in the negotiations, Austria, France and Germany returned to the negotiating table on the basis of a commitment that the Committee would issue an official statement assuring them on issues of concern. These statements are not legal acts, but they will provide some kind of assurance as the Commission will oversee the implementation of the Artificial Intelligence Act.
German Digital Minister Volker Wissing, who is the leading skeptic of the AI Act in the German Union, told his spokesperson"We ask the European Commission to clarify that the Artificial Intelligence Act does not apply to AI applications in medical devices. "
A statement by the European Commission circulated by Politico among EU diplomats before the vote revealed plans to create a group made up of the authorities of EU member states"Group of Experts"。The functions of the group will be:"Advice and assistance"The European Commission applies and implements the Artificial Intelligence Act.
In particular, the expert group will assist the European Commission in avoiding overlap between the Artificial Intelligence Act and other EU regulations, including the Medical Device Regulation, the In Vitro Diagnostic Devices Regulation and the Machinery Regulation.
The statement also obliges the ** Industrial Intelligence Office to be advanced"Universal"Provided by the developers of the AI model"Detailed guidance"to explain how to disclose a summary of copyrighted material used to train the software. The statement also reaffirmed the Commission's commitment to promoting innovation in the field of artificial intelligence and in updating certain texts affecting advanced artificial intelligence"Ensure a flexible and future-proof legal framework"。
Finally, the statement allows member states to adopt more restrictive rules and safeguards on technologies such as facial recognition, emotion recognition, and biometric classification.
The Artificial Intelligence Act still needs to be formally approved by the European Parliament. The text will be endorsed at the Committee level in two weeks' time and is expected to be voted on in plenary in April.
Disgruntled pro-privacy legislators may still try to block progress on the law by proposing amendments – which would require additional negotiations with the Council if they pass. But most of those involved in the drafting of the AI law in parliament are confident that it will be passed without any changes.