This article is a part of Poland Unpacked. Weekly intelligence for decision-makers
The European Commission has agreed to changes to the AI Act under the Digital Omnibus package, aimed at simplifying EU rules and reducing the burden on companies. The revisions are intended to facilitate the deployment of artificial intelligence while maintaining safety standards. Businesses and experts have broadly welcomed the direction of the changes.
In recent days, the European Commission announced a political agreement between the European Parliament and the Council of the EU on simplifying the implementation of the Artificial Intelligence Act (AI Act) as part of the Digital Omnibus package. The goal of the reforms is, among other things, to reduce administrative burdens on companies and strengthen the competitiveness of Europe’s technology market.
The new rules clarify regulatory obligations and simplify how they are applied by companies developing and deploying AI. Several of the facilitations have also been extended to mid-cap companies, while the relationship between the AI Act and other EU regulations – such as those on product safety – has been clarified to limit overlaps in legal requirements. Access to regulatory sandboxes, allowing AI to be tested in controlled environments, will also be expanded.
A timetable for the implementation of the provisions has also been agreed.
EC: Simplifying the AI Act to support innovation and security
Henna Virkkunen, Executive Vice-President of the European Commission for Technological Sovereignty, Security and Democracy, said that businesses and citizens expect two things from AI regulation: the ability to innovate and a sense of security. She added that the agreement reached meets both conditions. Thanks to simpler and more innovation-friendly rules, it facilitates the introduction of new solutions without lowering safety standards.
“Good progress on the Artificial Intelligence Act – AI-generated content will be labelled still this year, and regulatory sandboxes will also be established at the EU-wide level. We are now working on simplifications within the digital package. Step by step, towards a coherent and business-friendly body of EU law,” commented Dariusz Standerski, Deputy Minister of Digital Affairs, as quoted in a statement published on the ministry’s website.
Manager's perspective
Asseco: AI between security and business efficiency
From the perspective of Asseco and other technology companies that rely on data to develop their products, the upcoming Data Omnibus will be crucial. This legislative package is expected to simplify and reorganize rules governing data use. In practice, it would allow companies – not only technology firms but also banks, for example – to use existing datasets to train models under defined conditions, without having to seek consent on each individual occasion. This must, of course, be done in full compliance with security standards, ensuring that user data does not fall into the wrong hands or get passed on by AI systems.
It is also worth highlighting a shift in the approach to digital solutions known as “safety components,” which are currently classified as high-risk. In this area, the entry into force of certain rules (Annex III) has been postponed, while the regulatory framework is being simplified and limited to elements that genuinely affect safety.
For example, a credit scoring system used to assess clients, currently classified as high-risk, relies on AI. However, if the final decision is taken by a human advisor, the system may be excluded from that classification. Another example is HR systems used for initial screening of job applications. AI supports decision-making by analyzing data and generating recommendations, but it does not operate fully autonomously.
This approach reduces the risk of algorithmic decisions that could be harmful, as every outcome is ultimately reviewed by a human. Under the proposed rules, it also lowers compliance costs for this category of solutions. At Asseco, software development has always followed the principle of “human in the loop.”
AI Act in the face of the global technology race
Why is the EU simplifying regulations, and what do they mean for the business sector?
Ireneusz Piecuch, partner at the law firm DGTL, notes that when the AI Act was being introduced in mid-2024, there was broad consensus on the need for human oversight of artificial intelligence. However, as he stresses, in the ambition to position itself as a global leader in AI regulation, a key aspect was overlooked: artificial intelligence is not just another technology, but a “game changer.” In his view, the entity that wins the AI race could gain a durable advantage and, in practice, dominate the global market for years.
Neither the United States – the clear leader in the development of AI systems – nor China, which is competing for global leadership in this field, has opted for similarly stringent regulation. As Mr. Piecuch observes, after a period of what he describes as EU self-congratulation over its regulatory approach, the bloc has ended up in a situation of having ambitious assumptions that are difficult to translate into the realities of global competition.
A May 7, 2026 communication from the European Commission and the European Parliament confirmed an agreement on the scope of the changes. Work is now underway on the final wording of the document, which is expected to be adopted in the coming weeks.
It has also been acknowledged that implementing the new rules within the originally envisaged timelines is unrealistic.
Package of changes: timelines, bans and regulatory sandboxes
Ireneusz Piecuch outlines the key agreements:
- Timeline shifts. The Omnibus package “stretches out” the implementation schedule for high-risk systems. The key deadlines are now December 2027 (for biometrics, education, and HR) and August 2028 (for regulated products such as machinery and elevators). Companies gain nearly a year to prepare technically.
- Sectoral harmonization. This is a major shift for industry. Previously, a lift manufacturer with embedded AI would have had to comply separately with the AI Act and the Machinery Directive. The Omnibus removes this duplication. Conformity assessment procedures are now integrated, and technical standards are harmonized, reducing certification costs.
- New, specific prohibitions. In response to the rapid development of image-generation tools, the Omnibus tightens the list of prohibited systems. It explicitly adds a ban on applications generating intimate content without consent (so-called AI “nudification”) as well as systems producing material depicting child exploitation. The AI Act referred to these issues in general terms; the Omnibus spells them out explicitly.
- European sandbox. The AI Act originally envisaged regulatory sandboxes primarily at national level. The Omnibus introduces a centralized, EU-wide sandbox. This means that innovators from smaller countries that struggle to build their own infrastructure can test solutions directly under the supervision of experts in Brussels.
Manager's perspective
Comarch: Europe regulates AI – at what cost to competitiveness?
In the competition over the “raw power” of algorithms and the pace of their commercialization, US and Asian markets remain more flexible. EU regulations impose higher requirements on developers, which inevitably generates costs and, at least initially, slows down time-to-market. On the other hand, they build a foundation often missing among other global players: trust, transparency, and security.
Regulation should be designed with close attention to its impact on business and on Europe’s technological development. Otherwise, instead of fostering responsible innovation, it may become a barrier that slows the EU down in the global technology race.
Facilitations for business, but growing regulation in Poland
In addition, the scope of privileges has been extended to “small mid-caps.” The original AI Act provided facilitations (lower fees, simpler procedures) mainly for small and medium-sized enterprises (SMEs). The Omnibus now extends these benefits to small mid-cap companies, defined as firms employing up to 499 people. This is a nod to European scale-ups that have outgrown startup status but still need support to compete with US and Chinese giants.
“Interestingly, the finalization of the AI Omnibus coincided with work on Poland’s draft law on artificial intelligence systems. The long-delayed proposal has finally reached parliament, and its provisions are unlikely to be warmly received by Polish businesses. On top of the AI Act (over 100 articles and 140 pages), another regulatory instrument will be added (122 articles and around 100 pages), bearing all the hallmarks of “gold plating” – a term used to describe the creative expansion of EU legislation and the creation of additional national-level regulatory burdens,” predicts Ireneusz Piecuch.
“Taking the goats out”: EU AI regulation and the logic of regulatory overload
Commenting on recent changes to EU artificial intelligence rules, Ireneusz Piecuch refers to an old Jewish parable about “taking the goats out of the house.” The story tells of a man who complains to a sage about the cramped conditions and chaos in his home. He is advised to… bring more animals inside. When the chaos becomes unbearable, the sage instructs him to remove the animals one by one. As a result, the man experiences relief and comes to see his living conditions as tolerable. The problem appears to disappear.
According to Mr. Piecuch, a similar mechanism can be observed today in European AI regulation. First, an extensive AI Act was adopted. Now, under the Digital Omnibus initiative, Brussels is attempting to partially reduce the regulatory burden it has created for companies.
Nevertheless, in his view, the direction of change should be assessed positively. At the same time, the current proposals may still prove insufficient to deliver a meaningful simplification of the regulatory environment for companies developing and deploying AI solutions.
Expert's perspective
Why Europe is falling behind despite its innovation potential
Historically, however, this has not been fully successful. Regulation was implemented through directives, and each member state transposed them differently, resulting in nearly 27 legal regimes across the EU.
More recently, this has changed. The EU now relies more on regulations rather than directives, which are meant to be applied uniformly. In this respect, the Union actually holds an advantage over the United States.
Regulation itself does not prevent European companies from competing – there is a great deal of innovation emerging in Europe. Nor does Europe face major barriers in commercializing scientific discoveries. The real weakness lies in scaling solutions, a process hampered by the lack of a truly unified investment market.
The United States wins primarily due to vast capital resources and long-standing support for its own technology sector, which reinforces its competitive edge. China operates in a similar way, heavily subsidizing selected industries over decades.
Europe is also the only continent that does not systematically favor domestic companies in public procurement. Countries such as the United States, China, India, Brazil, and Russia maintain various mechanisms that restrict access to public contracts for foreign firms. The EU also takes a fundamentally different approach to breaking up monopolies.
In generative artificial intelligence, a key question concerns copyright policy. In the United States, major companies chose not to respect existing copyright frameworks and broadly scraped the entire internet, including content from pirated sources, to build large language models. Europe’s approach differs significantly, which may indeed slow down the pace of AI development. However, this is less of an issue in other areas of AI, including machine learning and deep learning.
It is not the AI Act that explains why tools such as Bielik or Mistral are not widely used across Europe.
Key Takeaways
- The European Commission and the European Parliament have agreed on changes aimed at simplifying the implementation of the Artificial Intelligence Act under the Digital Omnibus package. The main objective of the reform is to reduce administrative burdens on companies and strengthen the competitiveness of Europe’s technology market. At the same time, regulatory obligations have been clarified and overlaps between different EU rules – particularly in the area of product safety – have been reduced.
- The changes are intended to make it easier for companies developing and deploying artificial intelligence by extending support to mid-sized enterprises and increasing access to regulatory sandboxes. In addition, the relationship between the AI Act and other EU regulations has been streamlined to limit duplication of obligations and improve the clarity of legal requirements. A key element is also the harmonization of sector-specific rules, as well as the extension of timelines for implementing parts of the regulation.
- The reforms also include a new approach to safety and prohibitions, including a more precise definition of banned AI applications and the introduction of a centralized EU-wide sandbox for testing AI solutions. At the same time, the article stresses that despite regulatory simplification, Europe continues to face the challenge of global technological competition, and the effectiveness of the new rules will depend on whether it succeeds in combining innovation with safety and a coherent internal market.