EU AI Office publishes draft template for the AI Act's fundamental rights impact assessment, and asks for feedback
The European Commission's AI Office has published a draft form, not yet adopted, for the assessment that Article 27 of the AI Act requires before certain high-risk AI systems are deployed: by bodies governed by public law and private entities providing public services (critical infrastructure excepted), and by deployers of high-risk AI for credit scoring and for life and health insurance pricing. The Commission says feedback is open until 15 January 2027.

The European Commission's AI Office published a draft template on 16 November 2026 for the fundamental rights impact assessment (FRIA) that Article 27 of the EU AI Act requires of certain deployers of high-risk AI systems. The Commission is collecting feedback until 15 January 2027. The 13-page document is a draft: its cover says "This draft has not been adopted or endorsed by the European Commission."
The Commission says the AI Office developed it "with input from the EU Agency for Fundamental Rights (FRA)". Background: our explainer on the risk tiers.
Who has to do a FRIA
Article 27(1) applies before deploying a high-risk AI system referred to in Article 6(2), "with the exception of high-risk AI systems intended to be used in the area listed in point 2 of Annex III", which is critical infrastructure. The draft notice lists the deployers covered:
- "bodies governed by public law, such as government bodies or public institutions", and private entities providing public services, when they deploy a system in an Annex III high-risk area, critical infrastructure excepted;
- deployers of systems "evaluating creditworthiness or establishing a credit score, except for AI systems intended for financial fraud detection";
- deployers of systems "used for risk assessment and pricing in life and health insurance".
The Act does not define "bodies governed by public law", the draft says. It suggests the concept "may be understood as covering State organs and regional and local authorities, including legislative, executive, administrative, law enforcement and judicial authorities". It says the concept may also cover other entities established under public law to perform public tasks, and EU institutions and bodies. For private entities, it cites Recital 96's examples: "education, healthcare, social services, housing [and the] administration of justice".
When, and the notification
The draft says the FRIA "needs to be conducted prior to the deployer's first deployment of a high-risk AI system", for every system and all its purposes. It must be updated if, for example, "the processes concerned, the persons affected, or the version of the system or its underlying model" change.
Article 27(3) says the deployer "shall notify the market surveillance authority of its results, submitting the filled-out template". The draft says this goes to the authority of the deployer's Member State, but the route is still a placeholder: "Deployers can do so by [link to be added later]." Where a market surveillance authority has exceptionally authorised the system under Article 46(1), deployers may be exempt from notifying, the draft says, but not from the assessment itself.
The draft says public authorities, EU institutions and bodies, and those acting on their behalf that deploy high-risk systems in certain areas will have to register a summary of the findings in the public EU database.
The Commission says deployers "will need to comply with this obligation as of 2 December 2027."
What the template asks
The template "follows the structure of Article 27(1) AI Act". It has seven numbered sections, then a submission date and signature, and the public summary page:
- General information: the organisation, a responsible person, EU database numbers and the Annex III area of use; whether this is a first deployment, an update or relies on an earlier assessment; and which stakeholders were consulted.
- Context of use: the intended purpose "as defined by the provider in the instructions for use", the processes or tasks, and the period and frequency of use.
- Likely affected persons or groups: 11 tick boxes, including children, older people, women, LGBTIQ+ persons, persons with disabilities, and third-country nationals including asylum seekers and refugees.
- Fundamental rights impact and risks of harm: first the right to the protection of personal data, cross-referring to any data protection impact assessment, then one block per further risk. Each block asks which of 27 listed rights (plus "Other") it affects, which groups, and its likelihood and severity from Low to High, where High means "highly probable or already observed in comparable deployments". Then come mitigating measures and the remaining risk.
- Human oversight: the measures in place, who carries out oversight, "what authority they have to override, disregard or reverse the system's output", and their training, including AI literacy.
- Measures if risks materialise: monitoring and response, "such as suspending the use of the system", and procedures for complaints, internal governance and redress.
- Overall conclusion: a short synthesis of the risks. The draft says "A function to generate this synthesis automatically will be considered for the automated tool."
Once adopted, the Commission says, the template will also be an automated tool on the AI Act Single Information Platform.
Testing the draft
The AI Office is also looking for deployers that will be subject to Article 27 to test the template on one of their own systems, with expressions of interest by 4 December 2026. Depending on demand, the Commission says, the AI Office "may have to select the cases for the testing". Participation "is voluntary and without prejudice to obligations under the AI Act."
For UK readers
UK government guidance published on 4 November says the Act applies in Northern Ireland "in a specific and limited way under the Windsor Framework from 10 November 2026". As we reported, most deployer requirements are out of scope there; the guidance says the exception is that deployers of high-risk systems must monitor them and inform providers of serious incidents, malfunctions or risks (our 5 November piece).
Why it matters
On our reading, this draft shows the form these deployers would complete and send to a regulator. The Commission says deployers will need to use it "Once the template is finalised", and the document calls its views "the preliminary views of the Commission services", including its reading of who counts as a public body. Its notification link is still to be added.
- European Commission: Commission seeks feedback on draft template for the fundamental rights impact assessment under the AI Act (16 November 2026)
- Draft Explanatory Notice and template for the fundamental rights impact assessment under Article 27(5) of the AI Act (PDF, 13 pages, draft)
- Regulation (EU) 2024/1689 (AI Act), Article 27 and Annex III
Ask Relay — he reads every question himself and replies personally by email.
