The European Union is trying to make its landmark AI regulation easier to apply just as companies rush to prepare for major compliance deadlines. Through a new “AI simplification” package under the Digital Omnibus, the European Commission says it wants to reduce administrative burdens for businesses while preserving the core protections of the AI Act, especially around safety and fundamental rights. That balance is now at the center of the region’s evolving regulatory strategy.
The timing matters. The AI Act is already in force, parts of it are already applicable, and the most significant compliance milestones are approaching quickly. For many companies, this has created a compliance scramble: internal teams are mapping systems, classifying risk, and building governance processes, even as the EU refines timelines, publishes guidance, and expands support tools. The result is not a rollback of the AI Act, but a more flexible attempt to make implementation workable in practice.
A simplification push without abandoning the AI Act
The European Commission has framed the new package as a practical adjustment rather than a political retreat. In its 7 May 2026 announcement, it presented the AI simplification package as part of the broader Digital Omnibus and the EU’s wider simplification agenda designed to boost competitiveness. That positioning is important because it shows Brussels is trying to respond to business concerns without dismantling the architecture of the bloc’s first comprehensive AI law.
Officials have been careful in their language. The Commission says the aim is to make AI Act implementation easier for EU businesses while maintaining the benefits the regulation is meant to deliver for society, safety, and fundamental rights. In other words, simplification is being sold as a method of smoother compliance, not a dilution of the law’s central safeguards.
This distinction matters for market participants. Companies that hoped simplification might mean fewer obligations altogether are instead seeing a more nuanced message: the EU still expects compliance, but it recognizes that the path toward compliance needs clearer instructions, better sequencing, and more usable support mechanisms. That is why the current debate is less about rewriting the AI Act than about making its rollout more manageable.
The AI Act timeline is still intact, but more flexible
The core dates of the AI Act remain in place. The Commission says the law entered into force on 1 August 2024, which continues to anchor the staggered application of obligations across the next several years. That phased structure was always part of the regulation, but the simplification effort is now reshaping how some of the milestones will operate in practice.
The first wave of obligations already started on 2 February 2025, covering prohibitions, definitions, and AI literacy requirements. Later, on 2 August 2025, rules for general-purpose AI models became applicable, with governance rules and GPAI obligations entering into force as the EU builds out its regulatory institutions. These dates show that the AI Act is no longer a future framework; it is an active compliance regime.
The next major date remains 2 August 2026. According to the AI Act Service Desk, the majority of rules start applying then, and enforcement begins at that point, while additional parts continue to phase in later. The full rollout is still forecast for 2 August 2027. So while simplification may adjust the practical path for some obligations, the broader compliance calendar is still very much alive.
High-risk AI rules are at the heart of the change
The most notable adjustment concerns high-risk AI systems. Under the Digital Omnibus proposal, the application of high-risk AI rules is being linked more directly to the availability of support tools such as harmonised standards, common specifications, or Commission guidelines. This marks an important shift because it acknowledges a basic implementation problem: businesses need technical and legal guidance before they can confidently meet some of the law’s most demanding requirements.
That does not mean high-risk regulation is disappearing. Instead, the Commission appears to be introducing more flexibility into the timeline for those requirements, especially where compliance depends on interpretive tools that are still being developed. For businesses, this may offer temporary relief, but it also creates uncertainty because the final pace of some obligations now depends on whether those support materials are delivered on time.
There is also a distinction within the high-risk category itself. The Commission says that rules for high-risk AI embedded in regulated products have a later transition period and are set to apply from 2 August 2027. That extended runway may be especially relevant for manufacturers and firms operating in sectors where AI is integrated into products already covered by existing EU safety and product laws.
Transparency and GPAI duties are still moving a
While some high-risk requirements may become more flexible, other obligations remain firmly scheduled. The Commission’s AI Act materials indicate that transparency rules are still due to come into effect in August 2026. That means companies deploying certain AI systems will still need to prepare for disclosure-related duties on the original timetable, even as debate continues over other parts of the law.
General-purpose AI is also no longer on the horizon; it is already in scope. The Commission says GPAI model rules became applicable on 2 August 2025, and governance structures at the EU level are meant to be in place as part of that process. This is significant because many businesses have focused public attention on high-risk use cases, while GPAI obligations are becoming operational in parallel.
The GPAI Code of Practice is one of the tools designed to help here. The Commission describes it as practical guidance covering transparency, copyright, and safety and security obligations. Although voluntary, it gives model providers and related actors a concrete reference point as they work toward compliance in an area where technical, legal, and commercial questions remain fast-moving.
Brussels is building a compliance-support stack
A central part of the EU response to the compliance scramble is not just legal simplification, but operational support. The Commission has launched the AI Act Service Desk and a Single Information Platform that includes a Compliance Checker. These tools are meant to help organizations determine whether they have obligations under the law and what steps they may need to take next.
That support stack is a tacit acknowledgment of how complex the AI Act can be in practice. Even sophisticated companies often struggle with threshold questions: whether a system is covered, whether it falls into a high-risk category, what documentation is required, and when each obligation applies. By centralizing guidance and triage tools, the Commission is trying to reduce confusion before it becomes non-compliance.
The AI Office is also preparing guidelines intended to provide clear and practical instructions on how the AI Act should be applied alongside other EU rules. This is especially important for firms already dealing with product safety law, data protection law, consumer law, or sector-specific regulation. In that sense, the compliance-support stack is not just informational; it is becoming a core part of how the EU expects the law to be implemented.
Voluntary tools are becoming a bridge to readiness
The European Commission is also leaning on voluntary mechanisms to smooth the transition into the new regime. The AI Pact is one of the most visible examples. It is designed to support future implementation and encourage providers and deployers to begin aligning with the AI Act before every mandatory requirement fully applies.
According to the Commission, more than 230 companies had signed the AI Pact pledges. That level of participation suggests strong industry interest in early alignment, whether for risk management reasons, customer confidence, or regulatory signaling. In practical terms, companies may see voluntary engagement as a way to avoid being caught unprepared when binding deadlines arrive.
These tools also reveal an important feature of the EU’s current strategy. Brussels is not relying only on enforcement threats. It is trying to create a graduated pathway that combines formal obligations, soft-law guidance, self-assessment tools, and voluntary commitments. For companies in the middle of a compliance scramble, that mix may be more useful than a purely punitive model, even if it does not eliminate ambiguity.
Competitiveness, enforcement, and the moving target problem
The simplification debate is closely tied to the EU’s competitiveness agenda. The Commission has explicitly connected the package to the goal of boosting Europe’s business environment, arguing that implementation must be workable if the bloc wants to remain attractive for innovation and deployment. This reflects a broader recognition that regulatory ambition alone is not enough; execution matters just as much.
At the same time, enforcement structures remain central. The Commission says supervision and enforcement will involve Member States, national authorities, the European AI Office, the AI Board, and related bodies. So even with simplification, the AI Act remains a serious regulatory framework backed by a multilayered oversight model. Companies should not mistake procedural flexibility for an absence of scrutiny.
The main practical challenge is that some businesses still face a moving target. The Commission’s own materials suggest that the pace of certain high-risk obligations depends on whether harmonised standards, common specifications, or guidelines are available in time. That leaves compliance teams in a difficult position: they must prepare for obligations that are approaching fast, even though some of the tools needed to interpret those obligations are still in development.
That is why the phrase “EU simplifies AI rules amid compliance scramble” captures the current moment so well. The EU is not stepping back from the AI Act, which it still presents as the world’s first comprehensive AI law. Instead, it is trying to make a demanding framework more navigable just months before the main August 2026 deadline becomes the dominant focus for businesses across the region and beyond.
For companies, the message is clear but mixed. The direction of travel has not changed: compliance expectations are real, enforcement is coming, and the full rollout still points toward 2 August 2027. But the path is becoming more adaptive, with support tools, voluntary codes, and more flexible sequencing for some high-risk obligations. The scramble, in other words, is not over. It is simply being managed more actively by the institutions that created it.
