December 2023, European Commission President Ursula von der Leyen welcomed The newly completed EU Artificial Intelligence Law is set as a law that will “transform European values into a new era”. Three years later, after the same law was significantly revised, she described it as creating a “simple and innovation-friendly environment”.
Von der Leyen’s shift in rhetoric from values-driven governance to innovation-focused regulation reflects a broader shift in Brussels’ regulatory stance over the past three years. AI law has become one of the prime examples of that evolution.
Correction points Decided on May 7th It delayed implementation of rules governing AI systems in high-risk environments until 2027 or 2028 and removed industrial applications from the scope of these requirements entirely.
Importantly, however, the revisions do not change the timeline for the rule, which covers general-purpose AI models developed by companies such as OpenAI and Anthropic, and is scheduled to go into effect in August. As a result, some analysts and lawmakers argue that while the AI Act’s implementation timeline has been relaxed, its core regulatory framework remains in place.
“The fundamental purpose of this bill hasn’t changed as dramatically as some commentators say,” said Nicole Lemke, a senior policy researcher who studies AI systems at the European think tank Interface. “AI law has been, and continues to be, based on a risk-based architecture.”
What has changed in EU AI law?
When the AI Act was first approved in early 2024, it set ambitious deadlines to regulate high-risk AI systems deployed in areas such as employment, critical infrastructure, and biometrics. It also introduced rules for general purpose AI models that power systems like ChatGPT, Claude, and Mistral.
In the proposed amendment, Implementation deadline Regulations for high-risk AI systems have been postponed from August 2026 to December 2027, and requirements for AI integrated into products that are already subject to sector-specific regulations have been postponed to August 2028.
The extension will give the European Commission more time to finalize its guidance on identifying high-risk systems after several months of delay, while also giving companies a longer period of time to prepare for compliance.
Another big change concerns AI used in industrial machinery such as car manufacturing and car manufacturing. state-of-the-art lithography machine A company that makes semiconductor chips. Securing exemptions for industrial use was a priority. German Chancellor Friedrich Merzwanted to give Germany’s legacy industries the freedom to adopt AI.
As a result, companies in these sectors will only be subject to existing sector-specific regulations. According to Julie Gabriel, co-founder and CEO of eyreACT, a company that automates compliance with EU AI law, this carve-out could provide an early hint at how the EU’s approach to AI regulation will evolve in the future.
The revisions also prohibit the use of AI to generate child sexual abuse material and so-called “Nude” app One that creates images of non-consensual intimacy.
But some lawmakers say the EU could have gone further to simplify the law. Axel Bos (EPP, Delaware), the shadow rapporteur on the amendment, said the only substantive deregulatory measure in the package was an industrial carve-out. Beyond that, he argued, little has changed in the law and most amendments merely delay implementation deadlines.
Why change EU AI law now?
The broader question for policymakers and analysts is why the EU moved to amend the AI Act so soon after it was adopted in 2024. Supporters say the changes were necessary to increase Europe’s competitiveness in the global AI race. But critics say this theory has been pushed by lobbying efforts from big US tech companies and that the law is still largely unenforced.
Much of the push for simplification across the EU’s digital agenda can be traced back to its origins. 2024 Draghi Reportcalled on the EU to accelerate strategic technological innovation.
The message has since been embraced by influential technology lobbying groups, including AmCham EU, the American Chamber of Commerce in the EU, and the ACT Association, an association representing small and medium-sized technology companies.
a joint letter A report released by American and European industry groups in April 2026 says they are “shared.”[d] The European Commission’s commitment to clear, simple and innovation-friendly implementation of the AI Act. ”
Vicente Domecq, head of technology and digital policy for the European Conservatives and Reforms in the European Parliament and a supporter of the AI law, said further tightening of regulations would push Europe further away from the global AI race. “The challenges are not just safety and ethics,” he says. “But the real challenge has to be innovation and investment.”
However, some analysts dispute claims that the AI Act has stifled innovation. Ultimately, AI laws remain largely unenforced, said Risto Ouk, director of European policy and research at the Center of Life Institute in Brussels. No fines have been imposed under the law, but key provisions governing high-risk AI systems and general-purpose AI models have not yet taken effect.
Rather, the lack of AI innovation in Europe is more closely related to a lack of capital and market fragmentation, said Luca Ignac, interim director of AI at the Center for Future Generations.
“[The innovation] “The debate often ignores the fact that the EU’s influence lies in the regulatory framework,” he said. The so-called Brussels effect A word coined by Anu Bradford. “How Europe positions itself in this environment is also about setting governance standards that other countries can also participate in.”
On the contrary, critics of the AI law reform say this. Lobbying activities by US Big Tech It played a central role in ensuring delays. Domecq said he has discussed AI policy with representatives of several U.S. technology companies. “Of course,” they did their part, he said. “I received strong insights and recommendations from companies that have their own interests, but of course know the market.”
Gabrielle begged different. “The narrative that this whole AI law is bad for innovation has been promoted and paid for by American hyperscalers and big tech companies.”
High-risk AI rules stall due to delays in EU AI Secretariat
Another factor in the amendment was the failure of the EU AI Secretariat to produce important guidelines in time. The EU AI law was originally passed in 2024, but lawmakers left some of the regulations to the EU AI Office for final decision. Guidelines on how to determine whether your system should: classified as high risk It was just published by the EU AI Secretariat in May this year.
“I’ve been to several technical standards meetings where three-hour discussions revolved around the term ‘ethics.’ What does this mean?” Wouk said.
Without an agreed definition of what constitutes high risk systemthat part of the law is unenforceable. Gabriel said part of the problem stemmed from the bill’s original drafting process, which was done by non-technical lawmakers who didn’t understand the difficulty of coming up with complex technical standards from scratch.
Voss, the shadow rapporteur on AI simplification, agreed. “We don’t get into the technical details, so we don’t understand the complexities of complying with the law.”
The final challenge, according to Uuk and Ignac, is ensuring the AI office has enough support to complete the extensive technical work required to properly regulate the model. AI offices have traditionally struggled with: hire someone with sufficient experience Technical staff.
What’s next for EU AI legislation?
AI law has just undergone a fundamental overhaul, but the toughest challenge yet for the EU’s efforts to regulate AI will come in early August, when implementing regulations for a general-purpose AI model come into force.
Starting August 2nd, models offered by US companies such as OpenAI, Anthropic, and Google, as well as European companies such as Mistral, will code of practicea document defining transparency, copyright, and safety guidelines for the model was completed last year. If companies do not comply, the AI Bureau can order model recalls, impose fines and change mandates. This is the first regulatory obligation to cover all companies offering models within the EU.
Then there are questions about the long-term trajectory of the EU’s efforts to regulate rapidly evolving technologies. New AI applications, such as AI agents, can directly Subject to AI law
Vos said there is a sense that the AI Act is already “outdated” given the rapid advances in AI. He argued that EU institutions need to regularly review and update the framework to ensure it is responsive to new technological developments and new applications.
“The difficult question is whether Europe can build the institutional capacity to regulate technology at the pace of its development,” Ignac said. “This is the biggest challenge for us going forward, which is to build organizational capacity in a way that does not grow bureaucracy, but instead grows true technical expertise.”
sign up for of parliament weekly newsletter
Every Friday, our editorial team goes behind the headlines to provide insight and analysis on the key stories driving the EU agenda. Subscribe for free here.
