AI Act fundamental rights impact assessment: the Commission publishes the draft template

On 16 November 2026 the European Commission published the draft template questionnaire for the fundamental rights impact assessment required by Article 27 of the AI Act. It will have to be used by public bodies, private entities providing public services and anyone using high-risk AI to assess people’s creditworthiness or to price life and health insurance, before putting those systems into use. The obligation applies from 2 December 2027 and the consultation is open until 15 January 2027. We explain what the questionnaire asks and what to prepare now, including if you are the provider of the system.

Table of contents

What the Commission has published

The European Commission’s AI Office has published a draft explanatory notice and template questionnaire for the fundamental rights impact assessment (FRIA) that the AI Act requires from certain deployers of high-risk AI systems. Article 27(5) tasks the AI Office with developing this template, including through an automated tool, so that those organisations can comply in a simplified manner. The EU Agency for Fundamental Rights (FRA) contributed to it.

  • Public consultation: feedback can be sent until 15 January 2027 through an EU Survey form.
  • Testing the template: deployers subject to the obligation can apply by 4 December 2026 to test it on one of their own systems, with support sessions from the AI Office. Participation is voluntary and the AI Office will select the cases.
  • Final version: once adopted, the template will be set up as an automated tool on the AI Act Single Information Platform.

It is a draft. The Commission itself states that it has not been adopted and reflects the preliminary views of its services, so it may change.

Who has to carry out the assessment

Not every user of high-risk AI. Article 27(1) of Regulation (EU) 2024/1689 requires it, before deploying a high-risk AI system listed in Annex III, from three groups of deployers:

  • Bodies governed by public law. The Regulation does not define the concept. The explanatory notice suggests drawing on the case law of the Court of Justice and covering State organs, regional and local authorities and other entities established under public law to perform public tasks.
  • Private entities providing public services. The notice refers to Recital 96, which gives education, healthcare, social services, housing and the administration of justice as examples, and considers that other services of a similar public nature are also covered.
  • Anyone using AI to evaluate the creditworthiness of natural persons or establish their credit score, except systems used to detect financial fraud, and anyone using it for risk assessment and pricing in life and health insurance (Annex III, point 5(b) and (c)). Here the obligation reaches private banks, lenders and insurers, whether or not they provide a public service.

In all cases, systems intended for critical infrastructure (Annex III, point 2) are excluded.

When it has to be done

Before the first deployment of the system, once for each system and covering all its purposes of use. If a relevant element changes later, such as the processes in which it is used, the persons affected or the version of the system or its underlying model, the assessment must be updated (Article 27(2)). According to the Commission, the obligation applies from 2 December 2027, the same date as the other obligations for Annex III high-risk AI systems.

What the questionnaire asks, section by section

The template follows the structure of Article 27(1) and has seven sections:

  1. General information: type of organisation, the Annex III area in which the system is used, whether it is the first deployment, whether the assessment relies on a previous one or on the provider’s, and which stakeholders were consulted.
  2. Context of use: the intended purpose according to the provider’s instructions, the processes or tasks in which the system is used, how its outputs are used, and for how long and how often.
  3. Persons and groups affected: a checklist including children, older people, women, persons with disabilities, persons in a vulnerable social or economic situation, minorities and third-country nationals, among others.
  4. Risks to fundamental rights: first data protection, cross-referring to the GDPR data protection impact assessment where one exists; then, risk by risk, which rights are affected, which groups, with what likelihood and severity (low, medium or high), which measures reduce it and what risk remains afterwards.
  5. Human oversight: which measures from the instructions for use have been implemented, who carries out the oversight, what authority they have to override or reverse the system’s output and what training they have received.
  6. Measures if the risk materialises: how it is detected, whether use is suspended, the reporting to the provider required by Article 26(5), the handling of complaints from affected persons and internal governance.
  7. Conclusion: a summary of the risks identified.

Public authorities that register the system in the EU database also fill in a summary that will be published there (Article 49(3)). Once completed, the template is attached to the notification to the market surveillance authority (Article 27(3)). A deployer using a system exceptionally authorised by that authority, for example for reasons of public security (Article 46(1)), may be exempt from notifying, but not from carrying out the assessment.

How it fits with the GDPR impact assessment

It does not replace it: it complements it. Article 27(4) provides that, where an obligation is already met through the data protection impact assessment under Article 35 GDPR, the fundamental rights assessment complements it. The template puts this into practice: in the data protection section it allows deployers to cross-refer to the DPIA or to include its relevant parts. The DPIA looks at the risks of the processing for data subjects; the AI Act assessment looks at the risks of using the system for all fundamental rights, such as non-discrimination or the right to an effective remedy. We explain this in more detail in our article on the AI Act and the GDPR.

If you are the provider of the system

The assessment belongs to the deployer, but it depends heavily on the provider. It must start from the information in the instructions for use under Article 13, which in turn reflect the risk management system under Article 9. According to the Commission’s notice, that system covers the risks the provider can reduce through the design of the system, while the Article 27 assessment captures those arising from the specific context of use.

Article 27(2) allows the deployer to rely on existing impact assessments carried out by the provider. The Commission clarifies that the provider is not obliged to share them, but that doing so is good practice. For a tech company selling credit scoring systems to financial institutions or tools to public administrations, preparing that information will also be a commercial argument, and the contract should state what it delivers and what the customer may use it for. We cover this on our SaaS contract lawyer page.

Practical example: a lending platform

This example is fictitious. Credia is a consumer lending platform based in Valencia that decides applications with a credit scoring system licensed from a technology provider. It is the deployer of a high-risk system under Annex III, point 5(b), so before 2 December 2027 it will have to assess its impact on fundamental rights and notify the result. Credia asks the provider for the instructions for use and information on its risk management system, reuses its GDPR impact assessment for the data protection section, identifies older people and people with irregular income as affected groups, and defines who reviews rejected applications and how an applicant can complain. As the draft may change, it decides to test it on a real case and send its comments before 15 January.

What to do now

  • Check whether your organisation falls into any of the three groups and which Annex III systems it uses or plans to use.
  • Draw up an inventory of those systems and their purposes of use.
  • Ask the provider for the instructions for use, the risk management information and, if available, its own impact assessment.
  • Locate the GDPR impact assessments that already exist for that processing.
  • Appoint a team combining legal, technical and operational knowledge of the process in which the system is used, as the Commission’s notice recommends.
  • Consider testing the template, by 4 December 2026, or responding to the consultation, by 15 January 2027.

Which systems are high-risk and what other obligations they carry is explained in our AI Act compliance guide.

What is the AI Act fundamental rights impact assessment?

It is the analysis that Article 27 of the AI Act requires from certain deployers before using a high-risk AI system: how and when it will be used, who it may affect, what risks of harm it poses to their fundamental rights, what human oversight applies and what will be done if a risk materialises. The result is notified to the market surveillance authority.

Bodies governed by public law and private entities providing public services that deploy Annex III high-risk systems, except those for critical infrastructure, and anyone using AI to evaluate the creditworthiness of natural persons, except for fraud detection, or for risk assessment and pricing in life and health insurance.

According to the Commission, from 2 December 2027. It must be carried out before the system is first deployed and updated if a relevant element changes. The template is open for consultation until 15 January 2027 and may change before it is adopted.

No. It complements it (Article 27(4)). If you have already carried out a data protection impact assessment for that processing, the Commission’s template lets you cross-refer to it or include its relevant parts in the data protection section, but the other fundamental rights must also be assessed.

The obligation lies with the deployer, who knows the context of use. But the deployer can rely on impact assessments carried out by the provider where the case is similar (Article 27(2)). The provider is not obliged to share them, although the Commission considers it good practice.

The fundamental rights impact assessment arrives with more than a year to spare and with a template that can still be commented on. Now is a good time to identify the systems affected and gather the provider’s information, rather than waiting for the last quarter. At Innovatech we advise both those who develop and those who use AI systems on bringing them into line with the AI Act, as part of our AI Act legal services. Tell us about your case and we will give you a free initial assessment.

Managing Partner at Innovatech Legal | Website | + posts

Marta Suárez-Mansilla is Managing Partner of Innovatech Legal and a Spanish lawyer (abogada), Madrid Bar (ICAM), working in technology law. She completed Harvard Law School's Copyright course and BerkeleyX's Blockchain programme, and has advised technology companies for more than eight years.