The EU AI Act's Industrial Exemptions Are Narrower Than Manufacturers Think

When the European Union's AI Act entered its phased enforcement period in February 2025, many manufacturers assumed their industrial AI systems would fall under the regulation's broad exemptions for "safety components" already covered by existing sector-specific legislation. That assumption is proving dangerously wrong.

The European Commission's first round of interpretive guidance, published in January 2026, clarifies that AI systems used in industrial settings are only exempt from the AI Act if they are already regulated as safety components under the Machinery Regulation or the ATEX Directive. Systems that use AI for quality inspection, predictive maintenance, or production optimization — the vast majority of industrial AI deployments — are fully subject to AI Act requirements.

For high-risk applications, that means conformity assessments, technical documentation, human oversight requirements, and transparency obligations. The penalties for non-compliance can reach 3% of global annual turnover.

"Most manufacturers we talk to are 12 to 18 months behind on compliance planning," said Dr. Anna Googl, partner at Bird & Bird's technology practice in Munich. "They assumed industrial AI was carved out. It's not."

The practical impact is significant. Any AI system that influences a safety-related decision — from a quality inspection algorithm that determines whether a part ships to a predictive model that schedules equipment maintenance — may need to be classified as high-risk and documented accordingly.