Key takeaways
- Prohibitions, including emotion recognition at work, and AI literacy have applied since 2 February 2025. Check these first.
- The AI Omnibus (Regulation (EU) 2026/1744, in force 27 July 2026) moved high-risk dates to 2 December 2027 for Annex III and 2 August 2028 for Annex I.
- For most plants, the likeliest high-risk exposure is not machines but people: AI that allocates tasks, monitors or evaluates workers falls under Annex III employment.
- You become a provider, with much heavier duties, if you put your name on an AI system, change its intended purpose or substantially modify it.
- The Omnibus narrowed what counts as a safety component and moved machinery to a separate route, so check classification again even if you did it in 2025.
The EU AI Act (Regulation (EU) 2024/1689) entered into force on 1 August 2024 and has applied in stages since. In July 2026 it was amended by the "Digital Omnibus on AI" (Regulation (EU) 2026/1744), which moved the high-risk dates and changed several definitions that matter to manufacturers. This page sets out where things stand as of November 2026: which role you play, which tier your use cases fall into, and what to do now.
The timeline as of November 2026
| Date | What applies | Relevance to a plant |
|---|---|---|
| 1 Aug 2024 | Act enters into force | None directly |
| 2 Feb 2025 | Prohibited practices (Art. 5) and AI literacy (Art. 4) | Workplace emotion recognition banned; AI literacy measures expected |
| 2 Aug 2025 | General-purpose AI (GPAI) model obligations, governance, penalties | Applies to model providers, not to most manufacturers |
| 27 Jul 2026 | AI Omnibus (Reg. (EU) 2026/1744) in force | New dates and definitions below |
| 2 Aug 2026 | General application, incl. transparency duties (Art. 50) | Disclose AI interaction and label deepfakes where applicable |
| 2 Dec 2026 | New prohibitions (non-consensual intimate imagery, CSAM); end of grace period for machine-readable marking of AI-generated content placed on market before 2 Aug 2026 | Mainly relevant to GenAI providers |
| 2 Dec 2027 | High-risk requirements for Annex III systems | Employment and worker management AI |
| 2 Aug 2028 | High-risk requirements for AI in Annex I products | AI safety components in regulated products |
Note: The Omnibus was proposed by the Commission on 19 November 2025, agreed politically in May 2026, signed on 8 July 2026, published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. The final text uses fixed dates. The Commission's original idea of linking dates to the availability of standards was dropped. Dates checked against EUR-Lex and the Commission's AI Act page in November 2026.
Which role are you?
The Act puts most obligations on providers, who develop an AI system or have it developed and place it on the market or put it into service under their own name or trademark. Deployers are those using an AI system under their authority in a professional context. Importers and distributors have narrower duties.
For a typical manufacturer:
| Situation | Likely role |
|---|---|
| You buy a vision inspection system from an integrator and run it on your line | Deployer |
| Your maintenance team uses a SaaS predictive maintenance tool | Deployer |
| You build an in-house scheduling model and use it only in your plants | Provider and deployer (putting into service for own use) |
| You sell machines or products with AI features under your brand | Provider of that AI system |
| You rebrand an OEM's AI module under your name, change its intended purpose, or substantially modify it | You can become the provider (Art. 25) |
Watch out: "We just configure it" is not always true. Retraining a high-risk system on your own data, changing what it is used for, or shipping it under your nameplate can shift provider obligations onto you. Write the intended purpose into the purchase order and keep changes within it.
Where a product covered by certain EU product legislation (Annex I, Section A) contains a high-risk AI safety component, the product manufacturer who places the product on the market under its own name is treated as the provider of that AI. Machinery has a separate route since the Omnibus. See EU AI Act and the Machinery Regulation.
The risk tiers in plant terms
PROHIBITED banned (since 2 Feb 2025)
------------------------------------------
HIGH-RISK Annex I products / Annex III uses
------------------------------------------
TRANSPARENCY chatbots, deepfakes (Art. 50)
------------------------------------------
MINIMAL most plant AI: PdM, energy,
scheduling of orders, QC
Prohibited practices that touch the workplace
Several Article 5 prohibitions are relevant to manufacturers:
- Emotion recognition in the workplace, except for medical or safety reasons. A camera system that infers whether operators are frustrated or disengaged is banned. The Act's recitals distinguish physical states such as fatigue or pain from emotions, so a fatigue alert for drivers can fall outside the ban.
- Biometric categorization that infers sensitive characteristics such as race, political opinions, trade union membership or sexual orientation.
- Manipulative or deceptive techniques and exploitation of vulnerabilities that cause significant harm.
- Social scoring leading to unjustified detrimental treatment.
Penalties for prohibited practices reach up to EUR 35 million or 7% of worldwide annual turnover, whichever is higher. For SMEs the lower of the two amounts applies. The Omnibus extended a similar cap to small mid-cap enterprises for certain other categories of infringement.
High-risk route 1: Annex I products
AI is high-risk under Article 6(1) if it is a product, or a safety component of a product, covered by the EU harmonization legislation in Annex I and that product needs third-party conformity assessment. Examples in Section A include radio equipment, pressure equipment, equipment for explosive atmospheres, lifts, and medical devices.
The Omnibus narrowed this route in two ways relevant to factories:
- Safety component now means a component whose intended purpose is to prevent or mitigate risks to health and safety. AI used solely for non-safety aspects, such as user assistance, performance optimization, efficiency, automation or non-safety quality control, does not qualify. AI whose failure would endanger health and safety does.
- Machinery moved from Section A to Section B of Annex I. AI requirements for machinery will be set inside the Machinery Regulation rather than applied directly through the AI Act.
High-risk route 2: Annex III uses, especially employment
Annex III lists use areas that are high-risk regardless of product. For manufacturers, the critical one is employment, workers management and access to self-employment. It covers AI used for recruitment and selection, and AI used to make decisions affecting terms of work, promotion or termination, to allocate tasks based on individual behavior or personal traits, or to monitor and evaluate performance and behavior.
That catches more plant systems than people expect:
- Shift or task allocation that uses individual performance data.
- Productivity scoring of operators or pickers from MES, scanners or wearables.
- Automated CV screening for operator hiring.
Annex III also lists safety components in the management of critical infrastructure such as electricity, gas, heating and water supply. A manufacturer running its own utilities rarely falls in scope, but check if you operate supply for others.
An Annex III system can escape high-risk status under Article 6(3) if it does not pose a significant risk, for example because it performs a narrow procedural or preparatory task. A system that profiles natural persons is always high-risk. If you rely on this exemption as a provider, you must document the assessment and register the system.
What deployers of high-risk AI must do
From 2 December 2027 for Annex III systems, deployers must, among other duties:
- Use the system according to the provider's instructions for use.
- Assign human oversight to people with the competence, training and authority to do it.
- Ensure input data under their control is relevant and sufficiently representative.
- Monitor operation and inform the provider and authorities of serious incidents.
- Keep automatically generated logs under their control for at least six months, unless other law says otherwise.
- Inform workers' representatives and affected workers before putting a high-risk system into service at the workplace.
Most of these map onto controls you already know: work instructions, competence records, MOC and nonconformance handling. The worker information duty usually means works council involvement in Germany, Austria, the Netherlands and elsewhere, which takes time.
AI literacy (Article 4)
Article 4 has applied since 2 February 2025. The Omnibus replaced its wording: providers and deployers must take measures to support the development of AI literacy of their staff and others operating AI on their behalf. It no longer requires you to ensure a sufficient level for each individual. A proportionate response:
- A short briefing for everyone who uses AI tools, tied to your GenAI shop-floor policy.
- Deeper, role-based training for system owners, quality engineers who approve AI outputs, and anyone assigned human oversight.
- A record of who was trained on what.
General-purpose AI and copilots
GPAI model obligations have applied since 2 August 2025 to model providers. Models placed on the market before that date have until 2 August 2027. A manufacturer using a commercial copilot is a deployer of an AI system, not a GPAI provider. Two things still apply to you:
- Transparency (Art. 50) from 2 August 2026: if you deploy a chatbot that talks to customers or suppliers, people must know they are dealing with AI, and deepfake content must be disclosed.
- Contract hygiene: ask vendors how they meet their own obligations, and where your data goes. See AI vendor due diligence.
Effort for a mid-market manufacturer
| Task | Typical effort |
|---|---|
| Inventory and classify all AI uses | 5 to 15 person-days |
| Prohibited-practice check and fix | 1 to 3 days, unless you find a problem |
| AI literacy program | 3 to 10 days to set up, then onboarding |
| Each Annex III system as deployer | 5 to 20 days including works council |
| Becoming a provider of a high-risk system | Months; plan it as a product compliance project |
Common mistakes
- Relying on 2025 classifications. The Omnibus changed the safety component definition and the machinery route. Reclassify.
- Treating the delay as a pause. Prohibitions and AI literacy are already enforceable, and transparency applies from August 2026.
- Missing HR systems. Worker monitoring and scheduling tools bought by HR or operations are the most likely high-risk systems in a plant.
- Forgetting OEM AI. AI inside machines and SaaS tools still needs to be in the inventory.
Monday-morning checklist
- Build or update your AI use case inventory, including OEM and SaaS AI.
- Screen every entry against the workplace prohibitions. Switch off anything that infers emotions.
- Flag any system that allocates tasks, monitors or evaluates individual workers. Start works council conversations now.
- Record your role per system: deployer or provider.
- Set up role-based AI literacy measures and keep records.
- Re-check any "safety component" classification against the Omnibus definitions.
- Use a management system such as ISO/IEC 42001 to hold the evidence.
Run the AI Act risk triage assessment to get a first classification of your use cases.
Note: This is general information, not legal advice. Confirm your classification and obligations with counsel, using the consolidated text on EUR-Lex.
Sources and further reading
- Regulation (EU) 2024/1689 (Artificial Intelligence Act), EUR-Lex
- Regulation (EU) 2026/1744 (Digital Omnibus on AI), EUR-Lex
- AI Act: regulatory framework and application timeline (European Commission)
This guide is general information for manufacturers, not legal advice. Confirm regulatory obligations with qualified counsel.
FAQ
Frequently asked questions
When do the EU AI Act high-risk rules apply to manufacturers?
As of November 2026, after the AI Omnibus (Regulation (EU) 2026/1744), high-risk requirements apply from 2 December 2027 for stand-alone Annex III systems such as employment and worker management, and from 2 August 2028 for AI that is a safety component of products under Annex I. Prohibitions and AI literacy have applied since 2 February 2025.
Is AI visual quality inspection high-risk under the EU AI Act?
Usually not. Visual inspection for product quality is not on the Annex III list, and the AI Omnibus clarified that AI used only for non-safety purposes such as performance optimization or non-safety quality control is not a safety component. It can become high-risk if it performs a safety function in a regulated product, or if it is used to evaluate individual workers.
Can we use AI to detect worker fatigue or emotions on the line?
Emotion recognition in the workplace has been prohibited since 2 February 2025, except for medical or safety reasons. The Act's recitals treat detection of physical states such as fatigue differently from emotions, so a fatigue alert for forklift drivers can be lawful. The design matters, so get a legal review before deploying any camera or wearable that analyzes workers.
Do we have to train all staff on AI under Article 4?
Article 4 has applied since 2 February 2025. The AI Omnibus rewrote it so providers and deployers must take measures to support the AI literacy of staff who operate or use AI on their behalf, without guaranteeing a specific level for any individual. In practice, role-based training and a short policy briefing for AI users is a proportionate response.
Does using ChatGPT or Copilot make us a general-purpose AI provider?
No. GPAI model obligations fall on the companies that develop and place the models on the market. A manufacturer using a GenAI tool is normally a deployer of an AI system. You could take on provider duties if you build and market your own AI system on top of a model, or significantly modify a model, so check before productizing anything.
Related guides
Frameworks
EU AI Act and the Machinery Regulation: what machine builders and plants need to know
From 20 January 2027 the Machinery Regulation replaces the Machinery Directive and explicitly covers machine learning in safety functions.
Guide · 7 min read
Frameworks
ISO/IEC 42001 for manufacturers: an AI management system that fits your existing ISO stack
ISO/IEC 42001:2023 is the certifiable management system standard for AI.
Guide · 7 min read
Frameworks
NIST AI RMF for manufacturers: Govern, Map, Measure, Manage on the plant floor
The NIST AI Risk Management Framework is a free, voluntary playbook for managing AI risk.
Guide · 7 min read
AI Governance
How to build an AI use-case inventory in a manufacturing plant
An AI use-case inventory is a register of every AI system you build, buy or have embedded in equipment, with an owner, a risk tier and enough detail to act on.
Guide · 7 min read