EU AI Act for manufacturing
Manufacturing meets this Regulation through a different door and on a different date from almost every other sector. The main route is Annex I via existing product safety legislation at 2 August 2028, and the Digital Omnibus amended the Machinery Regulation directly, which most AI Act coverage did not mention.
Three routes in, and only one of them is the obvious one
| Route | What it catches | Date |
|---|---|---|
| Annex I — product Art. 6(1) | AI as a safety component of, or itself, a product under the Machinery Regulation or other Annex I legislation requiring third-party conformity assessment. Robot safety systems, collision avoidance, protective interlocks | 2 Aug 2028 |
| Annex III point 4 Art. 6(2) | Worker management, shift and task allocation based on individual behaviour or traits, productivity monitoring, performance evaluation. The route plants most often miss | 2 Dec 2027 |
| Annex III point 2 Art. 6(2) | Only where the plant is itself critical infrastructure, energy, water, gas, heating supply, and the AI is a safety component in its management and operation | 2 Dec 2027 |
Two different dates in one factory is the normal case, not the exception. A robot cell safety system is 2028; the system that decides who works which line is 2027.
Where this usually goes next
Three situations account for most people reading this page. Each has a different answer.
A deal is blocked on an AI questionnaire
Legal will not sign until you can evidence how AI is governed. HumanAudit’s AI Trust Package is a fixed $3,500 over five business days: a public trust page, a pre-filled SIG Lite / CAIQ / SSPA Section K questionnaire bank, and your AI inventory and classification.
You need ISO/IEC 42001 documentation
23 clause-mapped AIMS documents with all 38 Annex A controls pre-populated, editable and yours to keep, from $199. Or score your gaps first: 18 questions, free, no signup to begin.
You are not sure what reaches you
Twenty minutes with the founder. No prep, no deck, straight to the person accountable for the work. If none of this applies to you, you get told that on the call.
This reference is published by HumanAudit Inc. Not a law firm, not an accredited certification body, not a registered auditor. We build documentation, your counsel interprets it, and an accredited body of your choosing certifies you. How this is funded →
The safety component test does the work
Article 3 defines a safety component as a component fulfilling a safety function for a product or system, or one whose failure or malfunction endangers the health and safety of persons or property. Applied to a plant, that sorts most systems quickly.
| System | Likely position |
|---|---|
| Vision-based collision avoidance on a mobile robot | Safety component: Annex I |
| Protective interlock deciding when a guard may open | Safety component: Annex I |
| Visual quality inspection rejecting defective parts | Depends. Cosmetic defects, generally not. Safety-critical defects in a regulated product, arguably yes |
| Predictive maintenance scheduling servicing | Generally outside, unless failure itself endangers |
| Production scheduling and throughput optimisation | Generally outside |
| Shift allocation based on operator performance data | Annex III point 4(b) |
| Camera-based operator productivity monitoring | Annex III point 4(b): and check Art. 5 if it infers emotion |
Expert analysis. Our reading of common deployments against the definitions, not an authoritative classification.
What runs alongside
- Machinery Regulation conformity assessment. Where AI is inside a machinery product, the AI Act requirements are assessed within that existing procedure rather than a parallel one. Your notified body needs the relevant designation.
- Functional safety practice transfers substantially. Plants running established functional safety regimes already hold much of the Article 9 risk evidence, but oriented to hazard and failure rather than to fundamental rights, which Article 9 also requires.
- Article 26(7) applies before deployment. Employers must inform workers’ representatives and affected workers before putting a high-risk system into service at the workplace. In several member states works council co-determination binds earlier still. Employment guide →
- Article 5 in the plant. Emotion inference in the workplace is prohibited, except for medical or safety reasons. Fatigue and alertness monitoring is a genuinely arguable safety case; general mood or engagement monitoring is not. Article 5 →
Status labels on this page
Verified fact: The Art. 6(1) Annex I route and its 2 August 2028 date; the Machinery Regulation's presence in Annex I; that Reg. (EU) 2026/1744 amended Reg. (EU) 2023/1230 and Reg. (EU) 2018/1139 alongside the AI Act; the Art. 3 safety component definition; Annex III point 4(b); and Art. 26(7).
Expert analysis: The system-by-system placement table and the assessment of what functional safety practice transfers.
Unsettled: How the safety-component test will be applied to quality inspection systems in regulated products.
Two dates, one inventory
The practical first step is an inventory that records, per system, which route it takes and therefore which date applies. A plant with both Annex I and Annex III exposure cannot manage them on a single timeline.
Not sure where you sit?
The classifier maps your system against Articles 5, 6, 50 and Annex III. Twelve questions, no email.
Frequently asked
When does the EU AI Act apply to industrial machinery?
AI that is a safety component of, or is itself, a product covered by the Annex I harmonisation legislation and required to undergo third-party conformity assessment follows the Article 6(1) route and applies from 2 August 2028. The Machinery Regulation (EU) 2023/1230 is within that legislation. Notably, Regulation (EU) 2026/1744 amended the Machinery Regulation directly alongside the AI Act, so anyone tracking machinery compliance should read both.
Is predictive maintenance high-risk under the EU AI Act?
Generally not by itself. Predictive maintenance that schedules servicing is not usually a safety component fulfilling a safety function, and it is not usually an Annex III listed purpose. The analysis changes if the system's failure or malfunction would endanger the health and safety of persons or property, which is the Article 3 definition of a safety component, or if the system also allocates work to individual workers based on their behaviour or personal traits, which engages Annex III point 4.
Do factory worker monitoring systems fall under the EU AI Act?
Annex III point 4(b) covers AI intended to be used to make decisions affecting terms of work-related relationships, to allocate tasks based on individual behaviour or personal traits or characteristics, and to monitor and evaluate performance and behaviour. Systems doing that in an industrial setting fall within it, applying from 2 December 2027. Article 26(7) also requires employers to inform workers' representatives and affected workers before putting such a system into service.