EU Executive Commission, European Parliament, and Member State Representatives Discuss on 6th
"Consensus on Key Issues" Reported
Leading AI Regulation Possible with Legislation within the Year
Europe is reportedly on the verge of enacting the world's first 'Artificial Intelligence (AI) Regulation Law (The AI Act)'. The global spotlight is on what kind of regulations the EU will introduce as it aims to both protect regional competitiveness in the AI sector, which is being reshaped mainly by American companies, and prevent the actual harms caused by AI.
However, since there are conflicting interests among EU countries regarding various sectors linked to AI, such as the economy and defense, it is expected that coordination will be necessary for a considerable period even after the regulation is announced. Companies are closely watching whether the EU can complete the legislative process within this year as originally planned.
Major foreign media outlets, including Bloomberg News, cited multiple sources on the 6th (local time) reporting that after more than 10 hours of meetings in Brussels, Belgium, representatives from the EU Commission, the European Parliament, and delegations from the 27 EU member states appeared to have reached an agreement on the AI regulation bill. Although there has been no official announcement from the EU Commission or the European Parliament yet, it is known that the meeting participants found common ground on key issues after lengthy discussions.
On the previous day, the 5th, Carme Artigas, Spain’s Minister for AI and Digital Affairs, representing the EU presidency, stated that "85% of the wording of the bill has already been agreed upon."
According to the Associated Press, the EU’s AI regulation law is designed to ensure the safety of AI as a product, similar to automobiles, toys, and cosmetics. The law requires companies to document the data and functions used to train AI systems, prove that they have taken measures to mitigate potential risks, and allow external researchers to audit these measures. This means transparency in building and operating AI tools and systems.
Additionally, the EU plans to apply a 'risk-based' regulation to AI tools. Considering the characteristics of AI tools or systems, risks will be classified into four categories: unacceptable risk, high risk, limited risk, and minimal risk, with different regulations applied accordingly.
A representative AI tool classified as posing an unacceptable risk is real-time facial recognition systems. It is reported that lengthy debates were held on this issue during the discussions. Members of the European Parliament sought to ban its use entirely within the EU, while the EU Commission argued for partial allowance for national security purposes, leading to a clash between the two sides. The final agreement has not yet been disclosed.
Furthermore, the EU’s AI regulation law is expected to include provisions that consider AI tools creating 'deepfake' images or fake videos as low risk but require clear labeling that the content was produced by AI. It also includes penalties that could impose fines of up to 6% of global revenue for non-compliance.
If the EU finalizes and enacts the AI regulation law within this year, it will be the first AI regulation law established by a Western country. Amid the surge of generative AI technologies like ChatGPT this year and the intensifying competition among advanced countries to lead AI technology and regulation, the EU will take the lead in regulation.
The EU has pursued stringent AI regulations, considering the US’s failure to regulate big tech companies. Having first proposed the AI regulation law in 2021, the EU aims to complete the legislative process within this year and apply the law starting in 2026.
Especially with the European Parliament elections scheduled for June next year, if the law is not finalized by January next year, discussions will likely stall for some time. Within the EU, it is expected that re-discussion will only be possible 9 to 10 months after the election process concludes, prompting efforts to accelerate the legislation.
However, discussions have taken longer than expected due to differing interests among member states. Germany, France, and Italy, where AI technology development is rapid, voiced opposition to the EU’s AI regulation law by proposing some regulations be left to self-regulation. For example, France opposed the law because it planned to deploy AI-based smart cameras for security and public safety ahead of the 2024 Paris Olympics, which would be difficult under the new regulation.
Europe has solidified its leadership as a regulatory authority by continuously pressuring IT companies for years, including through the 2018 General Data Protection Regulation (GDPR). As a result, European rules are often adopted as global standards, leading to expectations that the AI regulation law will also be widely adopted worldwide upon introduction.
The United States, which needs to protect major AI companies such as OpenAI, Microsoft (MS), Google, and Amazon, has strongly opposed the EU’s AI regulation law. The US government and Congress view the EU’s approach as too harsh on the private sector developing technology and fear it could alienate AI developers. Led by Senate Majority Leader Chuck Schumer, the US is preparing its own AI regulation law, but according to The Washington Post (WP), it is lagging behind the EU in terms of speed.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
