Compliance & ContractsFree referencePDF, Excel, Word exportManufacturingEuropean Union

EU AI Act Risk Tier Reference for Manufacturers

The EU AI Act sorts AI systems by risk and assigns duties by role. Most ERP analytics land in the minimal tier, but a few plant and workforce uses do not. This table helps you see where each of your use cases is likely to sit.

Download:

Topic	Tier or role	What it means	Manufacturing example	Typical action
Prohibited practices (Article 5)	Unacceptable risk	Banned uses such as social scoring and emotion recognition in the workplace	Using camera AI to infer worker emotions on the line	Do not deploy; remove from any pilot
Annex III employment and worker management	High risk	AI used to recruit, evaluate, allocate tasks or monitor workers	AI that ranks operators by performance or allocates shifts based on behavior	Treat as high risk: risk management, data quality, logging, human oversight
Safety component of a regulated product (Annex I route)	High risk	AI that is a safety component of a product under listed EU product laws such as the Machinery Regulation	AI function that keeps a machine safe, built into the machine you sell	Follows the product conformity route; coordinate with product compliance
Critical infrastructure safety components (Annex III)	High risk	AI managing safety of critical digital or physical infrastructure	AI controlling safety functions for energy or water supply	Assess whether your plant or product falls inside the listed areas
Chatbots and AI interaction (Article 50)	Transparency obligations	People must be told they are interacting with AI	Customer-facing order status assistant	Show a clear AI disclosure in the interface
Synthetic content marking (Article 50)	Transparency obligations	AI-generated text, audio, image and video must be machine-detectable as artificial in many cases	Auto-generated marketing images or voice messages	Use tools that mark outputs; label deepfakes
Demand forecasting and inventory optimization	Minimal risk	No specific AI Act obligations beyond general law and AI literacy	ML safety stock suggestions in the ERP	Document use; apply normal governance
Predictive maintenance	Minimal risk (usually)	Not listed as high risk unless it is a safety component	Vibration model predicting bearing failure	Check if it acts as a safety function; if not, minimal risk
Visual quality inspection	Minimal risk (usually)	Defect detection on parts is not listed, unless part of a safety component	Camera model flagging solder defects	Keep records; escalate if used for safety decisions
ERP natural-language assistant	Minimal or limited risk	Transparency applies if people interact with it; high risk only if used for listed purposes	Chat over purchasing and inventory data	Disclose it is AI; avoid use for worker evaluation
AI literacy (Article 4)	Obligation on providers and deployers	Staff operating AI need sufficient AI literacy	Training for planners using AI suggestions	Keep a training record
General-purpose AI models	GPAI provider obligations	Duties fall on the model provider, with extra rules for systemic-risk models	A foundation model you download	Check provider documentation and your downstream duties
Provider role	Role	Develops an AI system or has one developed and places it on the market under its name	A manufacturer that fine-tunes and sells an AI feature in its product	Carries the heaviest obligations
Deployer role	Role	Uses an AI system under its authority in a professional capacity	A plant running a vendor AI quality tool	Follow instructions for use, assign human oversight, keep logs
Importer and distributor roles	Role	Bring in or make available AI systems from outside the EU	Reseller of non-EU inspection software	Verify provider conformity documentation
Date: entry into force	Timeline	Regulation entered into force on 1 August 2024	Not applicable	Use as the start of the phase-in
Date: prohibitions and AI literacy	Timeline	Apply from 2 February 2025	Not applicable	Confirm no prohibited use exists
Date: GPAI obligations	Timeline	Apply from 2 August 2025	Not applicable	Collect provider documentation
Date: most other provisions	Timeline	Scheduled from 2 August 2026, with Annex I product-embedded high risk from 2 August 2027; amendments proposing delays have been discussed	Not applicable	Check the current official timetable before planning

This is a simplified orientation for planning, not legal advice. Whether a use case is high risk depends on the exact wording of the Regulation and your role. Confirm with counsel and the official text on EUR-Lex.

AI use case inventory

An optional workshop to inventory your AI features and map them to risk tiers.

Optional. The tool and every export work without this. No spam.

How to use it

List every AI feature in your ERP, MES and plant systems, then find the closest row for each. Note your role: a company that uses a vendor tool is usually a deployer, while one that builds AI into a product it sells is a provider. Export the matching rows into your AI inventory and attach the typical action as a task for the owner.

Where manufacturers get surprised

The employment category is the common trap. AI that monitors or ranks workers, allocates tasks or evaluates performance is listed as high risk, even when it sits inside a scheduling or labor-reporting module. The other trap is the safety component route: AI embedded in a machine you place on the EU market can follow the product conformity rules, with extra dates.

Worker monitoring and ranking is the main internal risk
Embedded machine safety follows product law
Chat assistants need disclosure, not high-risk controls
Forecasting and maintenance are usually minimal risk

Common mistakes

Teams assume that open-weight or on-prem hosting avoids the Act; it does not, because the rules follow the use and the role. Others rely on a single date; the phase-in has several, and proposals to adjust timing have been discussed. Finally, many skip AI literacy, which is a simple obligation that is easy to evidence with a training log.

Frequently Asked Questions

Is an internal ERP chatbot high risk under the AI Act?

Not by default. A chat assistant over inventory or purchasing data is generally minimal or limited risk, with a disclosure duty if people interact with it. It becomes a concern if used for listed purposes such as evaluating workers.

Does the AI Act apply to non-EU manufacturers?

It can. The Regulation applies to providers placing systems on the EU market and to deployers in the EU, and also where the output of a system is used in the EU. Check your exposure with counsel.

What is the deployer duty for a vendor quality tool?

For high-risk systems, deployers must follow the instructions for use, assign competent human oversight, monitor operation, keep logs they control and inform affected workers where required.

Do I need to worry about dates already?

Yes for prohibitions and AI literacy, which already apply. Plan the rest against the official timetable and watch for amendments that may shift some dates.

Netray designs private ERP AI with logging, human approval and read-only access, which supports your own AI Act duties. We are not a law firm, so have counsel confirm classification.

Talk to an Expert Explore More Free Tools

Related Tools

Compliance & Contracts

More Compliance & Contracts tools

See all

Back to All Tools