EU negotiators soften AI regulation for industrial machinery, delaying high‑risk rules and narrowing scope
EU negotiators soften AI regulation for industrial machinery, delaying high‑risk rules until 2027 and exempting many machines to avoid double regulation.
The European Parliament and EU member states have agreed on a pared‑down approach to the EU AI regulation after months of talks, largely removing industrial machinery from the law’s scope and pushing back key compliance deadlines. The negotiated changes, reached in trilogue discussions, delay the start of obligations for certain high‑risk systems and create transitional rules for AI embedded in other products. Final approval by the Parliament and member states is still required before the revised AI regulation takes legal effect.
Trilogue negotiators agree to narrower scope
Negotiators from the European Parliament and the Council struck a compromise that limits the AI regulation’s immediate reach for many industrial machines. Lawmakers concluded that existing product safety rules, notably the Machinery Regulation, already cover a broad set of applications and that blanket inclusion would create overlapping requirements. Germany, speaking for several industrial capitals, had pressed for the carve‑out to avoid duplication and legal uncertainty for manufacturers.
The agreement was framed as a pragmatic adjustment to make the AI regulation more operable for industry without abandoning core safety goals. Officials said the move reflects a distinction between consumer‑facing AI systems and those integrated in industrial control and safety systems. Despite the narrowing, negotiators kept safeguards for systems deemed genuinely high risk.
Deadlines for high‑risk AI pushed back
Under the revised timetable, obligations for designated high‑risk AI systems will take effect in December 2027 rather than August 2026. The change is intended to give companies more time to adapt compliance processes, prepare documentation and align their products with the new rules. If an AI system is embedded within another product and not sold as a standalone item, the compliance deadline shifts further to August 2028.
The extended schedule applies to systems such as biometric face recognition, AI used in critical infrastructure and automated tools for credit scoring. Once in scope, providers will be required to implement risk management systems, maintain technical documentation and ensure training data meet quality and integrity standards, among other obligations.
Machinery largely placed outside immediate reach
The compromise explicitly recognizes that many industrial machines will fall under existing EU product rules rather than the AI regulation itself. That means AI functions tightly linked to machine safety — for example, safety interlocks, emergency shutdown logic or fail‑safe control loops — will often be governed by machinery and product liability frameworks. The intent is to avoid “double regulation” where companies would otherwise have to comply with two overlapping regulatory regimes.
Industry officials welcomed the clarification, saying it reduces the risk of conflicting requirements that could impede deployment of AI in manufacturing and automation. Regulators will still have authority to classify certain AI applications as high risk where safety or fundamental rights are at stake.
Industry voices relief and renewed caution
Business leaders immediately hailed the deal as a signal that Brussels is listening to industrial concerns, while stopping short of calling it a final victory. Hartmut Rauen, deputy chief executive of the German engineering association VDMA, said the outcome is positive but cautioned that companies must review the text carefully to verify that no new compliance traps were created. Trade groups emphasized the need for clear guidance to implement the boundary between product safety law and AI obligations.
The Siemens chief executive Roland Busch had warned publicly in recent weeks that overly burdensome rules could prompt investment and development to move outside Europe. That rhetoric underscored industry fears that a strict AI regime might erode competitiveness against U.S. and Asian rivals if regulatory costs multiply.
Lawmakers and trade bodies split on bureaucracy claim
Some lawmakers and industry representatives criticized the package as a modest compromise that still carries significant administrative burdens. MEP Svenja Hahn described the result as a “mini‑compromise with maxi‑bureaucracy,” arguing that further simplification is necessary for businesses across sectors. The comment reflects concerns that compliance tasks such as registration, documentation and auditing could weigh heavily on smaller manufacturers.
At the same time, Sarah Bäumchen, director of the ZVEI electrical industry association, said the trilogue outcome is an important first step to make the AI regulation workable for industrial deployment. She urged continued engagement to avoid duplicative rules that could stifle the growth of industrial AI within Europe.
Medical devices remain a contested area
Negotiators did not extend the machinery carve‑out to the medical devices sector, leaving questions about medical AI unresolved. Representatives warned that failure to align the AI regulation with the Medical Devices Regulation could perpetuate overlapping obligations for medical manufacturers that use AI for diagnostics, monitoring or decision support. Stakeholders called for a parallel revision of medical device rules to ensure regulatory coherence and patient safety.
Officials signaled the revision of the Medical Devices Regulation must remain a priority, arguing that harmonized definitions and clear interfaces between regimes are essential for innovation and regulatory clarity in health technology.
The trilogue agreement marks a pivotal moment in the EU’s attempt to balance safety, rights protection and industrial competitiveness in the AI regulation. Lawmakers must now vote on the amended text and member states must formally endorse it before the changes become law, and companies will be watching the final wording closely as they plan compliance, investment and product roadmaps.
