← All services

EU AI Act: what machinery manufacturers
must know as AI providers

If you place a machine with an AI component on the EU market, you are a provider — not a deployer. With all obligations that entails. Fines up to €35M.

Am I provider or deployer? Clarify now

Who is the provider?

Source: EU AI Act Art. 3(3)

Provider (Art. 3 EU AI Act)

Natural or legal person that develops or has developed an AI system and places it on the market or puts it into service under their own name or trademark.

For machinery builders: every machine with integrated AI delivered to customers makes your company a provider — not a deployer.

Does your company place a machine with an AI component on the EU market?

YES → You are PROVIDER → Art. 16 applies → 12 obligations

(even if a third party developed the AI)

NO → You only use a third-party machine?

You are DEPLOYER → Art. 26 applies → different obligations

When is machine AI high-risk?

Source: EU AI Act Annex I × Annex III

PathLegal basisMachinery example
Path 1Annex I: AI in products under EU product safety law (incl. MRO 2023/1230)Every machine with AI safety function under MRO → automatically high-risk
Path 2Annex III No. 2: AI controlling critical infrastructureAI-controlled process plants in chemicals, pharma, power generation

⚠ Not automatically high-risk:

Predictive maintenance, quality inspection, condition monitoring — when no safety function is involved. Still: transparency duties under Art. 13 remain.

The 12 provider obligations

Source: EU AI Act Art. 16 letters a–l

The correct path for machine AI

Critical note: no harmonised standards yet (June 2026)

Art. 40 EU AI Act allows conformity presumption via harmonised standards. For AI in machinery (Annex I × MRO 2023/1230) these standards do not exist as of June 2026.

Consequence: Annex VI self-assessment without standard reference has limited legal effect in supervision. Where MRO requires third-party assessment (Annexes IX, X), that also applies to the AI component (Art. 43(1) AI Act).

Recommendation: ISO/IEC 42001 as structural framework + contact notified body early.

→ Machinery Regulation 2023/1230

GPAI in machinery: Art. 51 ff. EU AI Act — separate duty set

GPT-based machine control, multimodal quality inspection (CLIP/vision-language models), or embedded foundation models fall under Art. 51 ff. EU AI Act — in addition to high-risk requirements.

Do I integrate a GPAI model (ChatGPT, Claude, Gemini, Llama…) as a component in my machine — or develop a purpose-built model? The answer determines whether Art. 51 ff. applies on top of Art. 16.

GDPR / DSG 2018 — when your machine processes personal data

SystemGDPR relevance
Vision systems (cameras, 3D sensors in work areas)Images and movement patterns of persons = personal data. Face recognition or biometrics: Art. 9 GDPR
ML performance monitoring of machine operatorsCycle times, error rates, intervention patterns per employee = personal data. Material impact: Art. 22 GDPR
Proximity sensors in cobotsLocation and movement patterns of persons = personal data, possibly biometric location
Voice control / NLP interfacesVoice recordings = biometric data under Art. 9 GDPR

Austria: DSB Vienna is the competent supervisory authority

Data Protection Authority Vienna (DSB) · www.dsb.gv.at · dsb@dsb.gv.at

Fines under GDPR / DSG 2018: up to €20M or 4% of worldwide annual turnover.

⛔ Austrian blocking condition: ArbVG §§ 96/96a

Any machine that captures employee behaviour, performance, location, or biometric characteristics needs a written works agreement before go-live under §96(1)(7) ArbVG.

Without that agreement, deployment is unlawful — regardless of CE marking and EU AI Act compliance.

→ Details: Workshop C includes ArbVG check

EU AI Act fines

€35M

Prohibited AI practices (Art. 5)

or 7% of annual turnover

€15M

High-risk violations

or 3%

€7.5M

Other violations

or 1%

OEM with €80M annual turnover + high-risk AI violation = €2.4M potential fine (3% EU AI Act). Prohibited practices (7%): €5.6M.

AI Act readiness for OEMs

  • AI component review: risk classification per Annex I / III
  • Technical documentation for AI component (Art. 11 + Annex IV)
  • Post-market monitoring concept for AI systems
  • Human oversight / kill-switch implementation (Art. 14)
  • Bias analysis for training data (Art. 10 — explicit duty)
  • Conformity path analysis (Annex VI vs third-party, Art. 43)

Request AI Act Readiness Check

We clarify provider vs deployer status, high-risk classification, and your conformity path for machine AI.

Request AI Act Readiness Check
Reply within 24 hours No commitment Confidential