EU AI Act: Big Tech pressure delays implementation
The EU is wrestling with the future of the AI Act. Between the aim of establishing a global standard for trustworthy AI and the pressure from Big Tech as well as the US government, which warn of innovation brakes, the Commission faces the question: How much regulation does AI need in Europe to avoid choking innovation while simultaneously protecting fundamental rights?
Source: Own Representation
Introduction
The AI Act, an EU regulation that has been in force since 1 August 2024 aims to establish a comprehensive legal framework for artificial intelligence. It is regarded as the world's first comprehensive AI law. The regulation is based on a risk-based approach that classifies AI applications into different categories: prohibited, high-risk, limited-risk and minimal-risk. For each category there are specific obligations ranging from complete bans to transparency requirements. Particularly relevant is that the AI Act also applies to providers outside the EU once their systems are used or offered on the European market, similar to the General Data Protection Regulation ( 3cl.org).
For so-called high-risk systems, which are used in healthcare, credit scoring, human resources or critical infrastructure, the AI Act imposes strict obligations. These include risk management, data governance, documentation, human oversight and ongoing monitoring ( eur-lex.europa.eu). In addition, there are special rules for large general-purpose AI models (General Purpose AI), such as powerful language models, that must meet particular transparency and safety requirements ( artificialintelligenceact.eu).
Currently there is discussion of a possible suspension or relaxation of the EU AI Act. This does not mean a complete abolition, but targeted adjustments such as additional transition periods, exemptions for tightly defined use cases and shifted penalties ( reuters.com). Specifically, it is discussed that certain high-risk applications would no longer have to be mandatorily registered in the central EU database if used only for internal or tightly limited purposes. It is also proposed that fines for violations in some areas would only take effect from August 2027 ( reuters.com). By Big Tech in this debate refer large, often US-based tech companies such as Meta, Alphabet, Apple, Microsoft, or Amazon, which invest heavily in generative AI and lobby in Brussels ( dig.watch).
The AI Act in detail
The path of the AI Act began in 2021 with a proposal from the European Commission and was officially adopted in 2024 after lengthy negotiations ( commission.europa.eu). Although in force since August 2024, it takes effect in stages. Prohibitions of unlawful practices already apply from February 2025, while large parts of the high-risk rules will only apply from August 2026 ( reuters.com).
Early in 2025, the Commission published guidelines explaining to employers, online platforms and police authorities which AI deployments are prohibited under the AI Act ( reuters.com). These include AI systems for monitoring employees' emotions, manipulative Dark Patterns in online shops, certain forms of social scoring and biometric-based predictions of crime. Violations can be fined up to 7 percent of global annual turnover ( simmons-simmons.com).
Source: eclear.com
The EU aims for comprehensive regulation of artificial intelligence with the AI Act.
At the same time, the EU Commission promotes its AI Continent program, which is intended to bundle investments in data centers, so-called AI Factories, and the training of skilled workers. A AI Act Service Desk to help companies implement the law. The action plan emphasizes that the AI Act aims to create a single market for AI and provide clear, predictable rules for all market participants ( europarl.europa.eu).
Development and pressure
At the same time, political pressure is growing. A leaked draft for a Digital Omnibus package shows that the Commission is considering exemptions and reliefs. These include an exemption from the registration requirement in the high-risk database for purely internal, tightly limited AI systems, a one-year transition period with no fines until August 2027 and flexible timelines for labeling AI-generated content ( reuters.com). Another Reuters report summarizes that the Commission is reacting to massive pressure from tech companies and the US government, warning of trade conflicts and competitive disadvantages ( reuters.com).
There is also opposition within Europe. The CEOs of Siemens and SAP publicly called for a revision of the EU data and AI laws, as overlapping and sometimes contradictory rules threaten Europe's digital competitiveness ( reuters.com). Lawmakers who helped negotiate the AI Act warn in Financial Times reports not to water down central safeguard mechanisms, as doing so could absolve powerful US corporations from accountability and facilitate the misuse of AI ( ft.com).
Experts in the European Parliament emphasise that the AI Act should be implemented through regulatory sandboxes, guidelines and standardisation in a way that does not hinder innovation and competitiveness. Civil rights organisations such as European Digital Rights (EDRi) warn, however, that simplification could be used as a door opener to weaken or delay core rights protection mechanisms ( europarl.europa.eu).
Source: YouTube
Analysis and Perspectives
The EU pursues two goals that are in tension: on the one hand it wants to set a global standard for trustworthy AI and safeguard citizens' fundamental rights ( digital-strategy.ec.europa.eu). On the other hand, Europe should become an attractive location for AI research and applications ( europarl.europa.eu).
From the perspective of Big Tech, it's about significant financial interests and control. The AI Act provides fines of up to 35 million euros or 7 percent of worldwide annual turnover for violations ( simmons-simmons.com). The extraterritorial reach of the regulation makes it a globally relevant standard ( 3cl.org). The influence of tech companies is massive: according to the Digital Watch Observatory, tech firms in Brussels spend around 151 million euros per year on lobbying ( dig.watch).
US industry associations like the US Chamber of Commerce have criticized the AI Act since 2023 as potential discrimination against US companies and as a risk to innovation ( uschamber.com). The US Vice President warned at the Paris AI Summit about excessive regulation in Europe ( apnews.com).
The systematic difference between the EU and the USA lies in the EU creating a uniform, legally binding regulatory framework, while the USA relies on guidelines, sectoral regulations and an AI.gov action plan that emphasises innovation and market forces but does not contain broadly comparable legislation ( 3cl.org, crsreports.congress.gov).
From an innovation perspective, the picture is ambivalent. Industry representatives such as Siemens and SAP lament a regulatory bottleneck ( reuters.com). Economists and legal experts, however, emphasize that a wisely implemented AI Act can strengthen innovation by building trust and providing legal clarity ( intereconomics.eu). A research paper argues that innovation that violates fundamental rights should not be considered progress ( arxiv.org).
Practical implications
For developers, startups and companies in Europe, the debate on a possible suspension of parts of the AI Act creates uncertainty. Additional transition periods and exemptions could provide short-term relief ( reuters.com). However, delaying the implementation of the AI Act would be risky, as essential parts are already in force and the Commission is working on concrete implementation ( digital-strategy.ec.europa.eu, europarl.europa.eu).
Source: ptechpartners.com
The EU AI Act will have wide-ranging practical effects on the development and deployment of AI systems.
Companies using AI should view the AI Act as a fixed framework with room to maneuver in the transition period. Independent information pages such as artificialintelligenceact.eu or guides from IBM explain the duties for different risk categories. These include documenting training data, human oversight and processes for handling errors.
For users, some protections of the AI Act are already a reality, such as the ban on manipulative Dark Patterns or certain forms of social scoring ( reuters.com). A delay or dilution of central parts of the law could mean that AI applications remain in operation longer without clear supervision. Critical analyses from organizations like EDRi, AlgorithmWatch or The Good Lobby are helpful here ( europarl.europa.eu, thegoodlobby.eu).
A simple check helps with categorization: First official sources ( EU-Webseite zum AI Act), then curated overviews ( AI-Act-Portale) and only then social media debates. This helps avoid sensational headlines.
Source: YouTube
Outlook
Open is how far the EU Commission will go with the relaxations: will only individual duties be stretched over time or will substantial safeguards be weakened so much that the AI Act loses its model function ( reuters.com)? Also unclear is the approval of the European Parliament and member states for a Digital Omnibus package that shifts painstakingly negotiated compromises ( ft.com).
Also remains open is the development of the transatlantic relationship. The US relies on innovation-driven regulation, while pressure for clear rules is growing there as well ( 3cl.org, crsreports.congress.gov). Whether a common framework emerges or two competing governance models remain is crucial for international AI providers.
Source: ve-ram.com
The infographic summarizes the key aspects of the EU AI Act, including its goals and the risk classification of AI systems.
Finally, the question is whether the EU keeps its promises: protect fundamental rights, enable innovation and make Europe a leading AI hub ( digital-strategy.ec.europa.eu). This depends on investments in infrastructure, education and open-source ecosystems, as well as the ability of supervisory authorities to respond quickly and wisely to new risks ( europarl.europa.eu).
The debate around the AI Act shows that Europe is still balancing fundamental rights protection and speed of innovation. The Commission's current proposals point to targeted relief and flexibility, not to abandoning the law ( reuters.com). The influence of Big Tech and the US government is clearly visible. It will hinge on whether European institutions keep their goals in sight ( dig.watch, uschamber.com).
For businesses and citizens, this means viewing the AI Act as a framework that is in a new phase. Whoever develops or uses AI should systematically check which risk category the applications fall into and what documentation and controls make sense. Citizens who pay attention to their rights should monitor official information, critical analyses and practical experiences. At this interface it will be decided whether Europe’s AI rules are a brake or a springboard.