EU rule-makers weigh a narrower AI Act as pressure builds from industry and governments

Brussels is working through a new compromise on the AI Act that would reduce duplication between the bloc’s AI rules and existing sector-specific laws, a sign that Europe’s landmark framework is shifting from broad principle-setting to the harder business of implementation. The talks, discussed on April 17, 2026, come as industry and several governments push for a lighter compliance path for industrial and high-risk AI deployments.

Brussels looks to trim overlapping AI obligations

The proposed compromise is aimed at one of the most practical complaints about the AI Act: companies operating in regulated sectors can end up navigating several layers of compliance at once. By easing the way the AI law interacts with sectoral regimes, negotiators are trying to cut redundant reporting and certification steps without reopening the core risk-based structure of the regulation.

That matters because the most burdensome parts of the AI Act are not the headline bans, but the obligations that will apply to high-risk systems used in areas such as infrastructure, employment, education, law enforcement and border management. For companies building or deploying those systems, even small changes in how the rules interlock can affect product timelines, vendor contracts and market entry plans.

Industrial AI is driving the pressure

The latest push is coming as European manufacturers and their political allies argue that industrial AI needs more room than consumer-facing applications. Germany’s Chancellor Friedrich Merz said on April 19, 2026, that industrial AI should face more regulatory freedom than other forms of AI if Europe wants to improve productivity, echoing complaints from major industry figures that the current regime could steer investment elsewhere.

That argument has become politically important because Europe is trying to balance two objectives at once: keeping the AI Act credible as a safety framework while avoiding rules that companies see as so complex they slow deployment. The compromise under discussion suggests lawmakers are now treating compliance friction as a competitiveness issue, not just an administrative one.

What the revision could change for developers

For AI developers and system integrators, the practical question is whether Brussels ends up clarifying responsibilities or merely shifting them around. A narrower interaction between the AI Act and sectoral laws could make conformity assessments more predictable, especially for firms selling into heavily regulated markets where software, hardware and safety rules already overlap.

Even without changing the law’s risk categories, that kind of revision would have commercial consequences. It could reduce the number of separate approvals or documentation streams companies need to maintain, and it could also affect which products are viable for rapid deployment inside the EU versus in faster-moving markets abroad.

The talks are not a final rewrite, but they show that the AI Act’s next phase is being shaped as much by implementation mechanics as by policy ambition. The question in Brussels is no longer whether the bloc should regulate AI, but how much friction it is willing to leave in the system.