Art.60
EU AI Act Guide › Chapter VI — Measures in Support of Innovation › Article 60

Article 60 – Testing of High-Risk AI Systems in Real World Conditions Outside AI Regulatory Sandboxes

High-Risk Systems SME Relevant ~2 min read · 476 words

Article 60 provides a pathway for testing high-risk AI systems directly in real-world conditions — outside the formal sandbox environment — before they are placed on the market. It is a structured exception that allows providers to gather real-world evidence of their system’s performance while operating under a controlled, authority-supervised testing plan.

⚠ Medium compliance impact for SMEs

WHAT THE ARTICLE IS ABOUT

A supervised pathway for real-world AI testing before market placement

Article 60 establishes the conditions under which providers of high-risk AI systems may conduct testing in real-world conditions outside of AI regulatory sandboxes — essentially field trials in live environments. This pathway recognises that laboratory testing and sandbox environments cannot always replicate the full complexity of real-world deployment, while ensuring such testing occurs under appropriate safeguards.

WHAT IT SAYS

Authority-approved plans, informed consent, registration, and time limits

  • Providers may conduct testing in real-world conditions outside sandboxes subject to a real-world testing plan approved by the relevant market surveillance authority
  • The testing plan must specify: the AI system description, objectives, duration, geographic scope, the categories of persons who will be subject to testing, and the safeguards in place
  • Testing must be registered in the EU database with a unique identification number before it begins
  • Informed consent must be obtained from persons who will be subject to the testing — with exceptions for law enforcement contexts
  • Testing must not last longer than necessary and is limited to the time specified in the approved plan
  • Persons subject to testing must not be adversely affected by the AI decisions made during testing — any decisions made by the system during testing are without legal effect unless the subject agrees otherwise
  • The provider must have adequate liability insurance covering potential harm to test subjects

WHO IS AFFECTED

Providers of high-risk AI conducting pre-market field testing

  • Providers of high-risk AI systems who need real-world data to validate their system before market placement
  • Market surveillance authorities who approve and monitor testing plans
  • Natural persons who are subjects of the real-world testing — they must provide informed consent
  • Organisations that host or facilitate real-world testing on their premises or with their users

WHAT IT MEANS FOR SMES

A structured way to gather real-world evidence before launch

  • If your high-risk AI system needs real-world validation data that a sandbox cannot provide, Article 60 gives you a formal pathway — but it requires upfront regulatory approval and careful planning
  • The informed consent requirement is non-negotiable: every person whose data or decisions are affected by your test must actively consent before the test begins
  • The no-adverse-effect rule is a strong protection for you: decisions made by your AI system during testing cannot be enforced against test subjects, limiting your liability exposure
  • The EU database registration requirement means your testing is publicly recorded — plan for this transparency in your communication with test participants
  • The liability insurance requirement is a real cost to factor in — check what your existing business insurance covers and whether you need specific AI testing liability coverage

Related Articles

← Back to the Guide
← Previous Art. 59 — Further Processing of Personal Data for Developing Certain AI Systems in the Public Interest in the AI Regulatory Sandbox Next → Art. 61 — Informed Consent to Participate in Testing in Real World Conditions Outside AI Regulatory Sandboxes