Skip to main content

Access to Essential Services

This page covers AI that supports or makes decisions about access to essential services such as credit, insurance, and benefits.

Why this is usually high-risk​

These decisions can materially affect housing, financial stability, health access, and other core life outcomes. Because the rights impact is substantial, this use is generally handled as high-risk candidate scope.

Fundamental Rights Impact Assessment before deployment (Art 27)

Deployers that are public bodies or provide public services, and creditworthiness / insurance-pricing deployers, MUST complete a Fundamental Rights Impact Assessment before deployment and notify the market surveillance authority of the result (Art 27).

Required obligations​

WHEN​

  • AI output influences approval, rejection, prioritization, or eligibility decisions for essential services.
  • Data sources, model behavior, or decision policy change in ways that may impact outcomes.

MUST​

  • Treat as high-risk candidate and escalate for formal classification.
  • Do not launch until required controls are approved for your role.
  • Check the Art 6(3) derogation: it may be lower-risk if it only performs a narrow procedural or preparatory task — but profiling of natural persons makes it always high-risk. Self-classifying as not-high-risk still requires a documented assessment (Art 6(4)) and EU-database registration (Art 49(2)).
  • Support the affected person's right to explanation for the decision (Art 86).
  • Re-run classification whenever decision logic, model, or data sources materially change.
note

AI used only for fraud detection in financial services, and for prudential capital-adequacy calculations, is generally treated as not high-risk — confirm the exact scope with legal/compliance before relying on this carve-out.

OUTPUT artifacts​

  • Documented feature scope and affected population record.
  • Formal classification and legal/compliance approval record.
  • Versioned change history for model/data/decision-policy changes.
Real-world illustration

Both decided under GDPR/ECHR, not the AI Act, but they map directly to essential-services scoring (Annex III area 5):

  • SyRI (Netherlands, 2020): a court struck down a government welfare-fraud risk-scoring system for breaching the right to privacy (Art 8 ECHR). (overview)
  • Dutch childcare-benefits scandal: a discriminatory fraud risk-selection method wrongly branded around 26,000 families as fraudsters; the Dutch DPA found it unlawful and the government resigned. (overview)