Fundamental Rights Impact Assessment (FRIA): who needs it and how to structure it
As of 18 November 2026 · checked against Regulation (EU) 2024/1689 and Regulation (EU) 2026/1744
Information only, not legal advice. The FRIA is a regulatory instrument with binding effect. Confirm scope and content with qualified counsel before submitting to the market surveillance authority.
Article 27 of Regulation (EU) 2024/1689 introduces the Fundamental Rights Impact Assessment, or FRIA. It is a deployer-side obligation specific to Annex III high-risk AI systems, distinct from the provider-side conformity assessment and from the GDPR DPIA, although it overlaps with both. Article 27 sits in Chapter III, Section 3, so it applies from 2 December 2027 after the Digital Omnibus, see the deadline timeline. This page sets out who has to produce a FRIA, the six elements Article 27(1) requires, and how the notification to the market surveillance authority works.
Who is in scope
The FRIA obligation applies to three groups of deployers of Annex III high-risk systems. Systems under Annex III, point 2 (critical infrastructure) are excluded in all cases:
- Bodies governed by public law. Any state authority, regional or municipal administration, or public-sector entity operating an Annex III system, regardless of size.
- Private entities providing public services. Recital 96 names tasks in the public interest such as education, healthcare, social services, housing and the administration of justice.
- Specific private operators in the financial sector. Deployers using systems for credit scoring (Annex III §5(b)) or for risk assessment and pricing of life and health insurance (Annex III §5(c)): irrespective of public or private status.
Private deployers outside these categories are not subject to Article 27 even when using Annex III systems. They may still face DPIA duties under Article 35 GDPR if personal data is involved.
The six required elements
| Section | Required content |
|---|---|
| (a) Processes | The deployer's processes in which the system will be used in line with its intended purpose |
| (b) Period and frequency | The period of time within which, and the frequency with which, the system is intended to be used |
| (c) Affected persons | Categories of natural persons and groups likely to be affected in the specific context |
| (d) Specific risks of harm | Risks likely to affect those categories, taking into account the provider's information under Article 13 |
| (e) Human oversight | How the human oversight measures are implemented, according to the instructions for use |
| (f) Measures if risks materialise | Including the arrangements for internal governance and complaint mechanisms |
Relationship to other assessments
A FRIA is not a DPIA, but the two often share evidence. Since the Digital Omnibus, Article 27(4) lets the deployer cross-reference the relevant sections of its DPIA, or take over relevant parts of it, where the DPIA already meets an obligation. A practical order:
- Run the DPIA first if personal data is involved: it scopes the data flows and lawful basis.
- Run the FRIA after the DPIA, reusing the data-flow diagram and adding the fundamental-rights lens beyond data protection (non-discrimination, dignity, access to justice, freedom of expression).
- Reference the provider's technical documentation under Article 11: do not duplicate it.
- If the system is procured, ask the provider for Article 13 instructions for use detailed enough to fill elements (d) and (e).
Working template structure
A working structure we suggest. Sections 6 and 7 go beyond what Article 27 requires:
1. Context
1.1 Deployer identity and competent authority
1.2 Annex III sub-area and provider reference
1.3 Decision being supported by the AI system
2. Process and use
2.1 Workflow diagram (where in the human process)
2.2 Volume and frequency
2.3 Duration and review period
3. Affected persons
3.1 Categories (applicants, beneficiaries, employees, ...)
3.2 Vulnerable sub-groups identified
3.3 Estimated population per category
4. Fundamental-rights risk assessment
4.1 Risk inventory (right -> scenario -> severity -> likelihood)
4.2 Discrimination test (protected characteristics)
4.3 Procedural-rights test (right to be heard, redress)
5. Mitigations
5.1 Human oversight (cross-reference Art. 14 measures)
5.2 Training of operators
5.3 Internal complaint-handling and external redress route
5.4 Logging and audit trail (cross-reference Art. 12)
6. Update triggers
6.1 Substantial modification of the system
6.2 Change of context of use
6.3 New evidence of disparate impact
6.4 Periodic review cadence (at least annually)
7. Notification
7.1 Filled-out AI Office template sent to the MSA
7.2 Date of submission and reference number Notification mechanics
- Once the assessment is done, the deployer notifies the market surveillance authority of its results and submits the filled-out template as part of the notification (Article 27(3)).
- The AI Office develops that template as a questionnaire, including an automated tool (Article 27(5) as amended by the Digital Omnibus).
- The obligation applies to the first use. The deployer may rely on earlier FRIAs or on existing assessments by the provider in similar cases, and must update the information when an element has changed (Article 27(2)).
- In the case of Article 46(1) (derogation for exceptional reasons of public security and similar), deployers may be exempt from notifying.
What we track as events
AI Act Radar emits events on the fria sub-topic when the AI Office or a national MSA publishes the notification template, sectoral guidance for credit scoring or insurance, or court decisions interpreting Article 27.
Sources
- Regulation (EU) 2024/1689, Article 27 and Annex III: EUR-Lex
- Regulation (EU) 2026/1744 (Digital Omnibus), amended Article 27(4) and (5) and Article 113: EUR-Lex
→ See live FRIA updates: https://aiactradar.com
→ Subscribe to webhook: https://docs.aiactradar.com/quickstart