EU AI Act Amendments Extend Compliance Deadlines and Narrow Scope

The recent amendments to the EU AI Act have sparked a debate on whether the changes represent a strategic adjustment or a step toward deregulation. The omnibus deal, agreed upon by EU lawmakers on May 7, pushes compliance deadlines for high-risk AI applications into 2027 and beyond, while narrowing the scope of existing obligations.

Key Changes to Compliance Deadlines

Under the new agreement, the deadline for compliance regarding AI used in employment, education, and health insurance has been extended from summer 2026 to December 2, 2027. Compliance for AI embedded in physical products, such as medical devices and industrial machinery, has been moved to August 2028. Notably, machinery has been entirely exempted from the AI Act, with its governance shifting to sector-specific regulations.

These amendments benefit small and medium-sized enterprises (SMEs) and small mid-cap companies by providing simplified technical documentation requirements and access to regulatory sandboxes. These changes aim to ease the burden on smaller firms testing AI systems under relaxed rules, a relief welcomed by manufacturers but met with skepticism from consumer advocates.

Diverging Regulatory Landscapes: EU vs. UK

For UK AI vendors looking to operate in the EU, the implications of the omnibus deal are significant. The extended compliance timelines grant an additional 16 months for high-risk obligations, while industrial AI vendors will face a different regulatory framework. This divergence raises challenges, as UK firms cannot assume that compliance in one jurisdiction will suffice for the other, particularly in sectors like financial services where regulatory scrutiny is intensifying.

As the UK Financial Conduct Authority (FCA) reevaluates its approach to AI regulation, asking for evidence-based compliance rather than theoretical justifications, it contrasts sharply with the EU's trend toward simplification. The Commons Science Committee's push to include generative AI within the scope of the Online Safety Act further illustrates the UK's regulatory evolution.

Political Pressures and Future Implications

The political dynamics surrounding the EU AI Act amendments reveal a landscape marked by lobbying and contrasting viewpoints. Companies such as Siemens and ASML advocated for the machinery carve-out, while Green MEP Sergey Lagodinsky raised concerns that this exclusion could lead to fragmented AI regulation across Europe. The decision to revisit the AI Act only two years after its initial implementation underscores the evolving nature of AI regulations and the complexities involved in balancing innovation with oversight.

Formal approval of the omnibus deal is anticipated in the coming months, with the new prohibition on non-consensual AI-generated sexually explicit images set to take effect on December 2. Lagodinsky's warning that legislative processes can lag behind innovation raises questions about whether the EU will effectively fill regulatory gaps through guidance and codes of conduct. For UK regulators observing from outside the EU, the possibility of swift non-legislative enforcement may emerge as a critical benchmark, overshadowing the significance of the new deadlines.

As the regulatory landscape shifts, stakeholders in both the EU and UK must navigate these changes carefully. The potential for divergence in compliance requirements could reshape how companies approach their AI strategies in different markets, highlighting the need to stay informed and adaptable in an era of rapid technological advancement.