Putting into Service
Article 3(11) EU AI Act — first use by the deployer
Practitioner's note: This article is practice-oriented compliance documentation, not legal advice. We are a compliance specialist, not a law firm. For legally binding information please consult a licensed lawyer.
TL;DR
Putting into service is defined in Article 3(11) EU AI Act: the first use of an AI system by the deployer, NOT test use, but productive use for the intended purpose.
What is putting into service?
Key consequences of putting into service:
- Deployer obligations under Article 26 EU AI Act apply
- For high-risk AI: the FRIA must be completed BEFORE putting into service (Article 27)
- AI literacy training for the deployer's personnel (Article 4) becomes mandatory
Practical example
A law firm internally tests an AI contract analysis tool (sandbox). Putting into service occurs only when the AI is used in the productive workflow for actual client matters.
Frequently asked questions
Is a pilot sufficient?
A pilot with real data + real use cases = putting into service. Sandbox testing without real data: no.
What must be completed BEFORE putting into service?
FRIA (for high-risk AI), AI literacy training, review of provider documentation, update of the privacy notice.
Who is liable?
The deployer for putting-into-service obligations, the provider for design/safety defects of the model.