Zum Hauptinhalt springen
AI Compliance Starter Kit · Template

Fundamental Rights Impact Assessment (Art. 27)

Updated on 2 min Reviewed by: Cenedril Editorial
EU AI Act Art. 27ISO/IEC 42001 A.5

An insurer uses an AI system that assesses applications for life and health insurance. An error hits above all the people who are refused a contract or offered one at a higher price. The fundamental rights impact assessment checks before deployment which groups are affected, which harms are possible and which oversight and remedies are in place.

Who is the template for?

For deployers carrying out a fundamental rights impact assessment as a legal duty, and for any organisation with an AI management system under ISO/IEC 42001 that wants to keep its impact assessments in a uniform way. The groups, impact areas and societal fields are listed as choices in the template.

Template: Fundamental Rights Impact Assessment

Full policy text

Fundamental Rights Impact Assessment

Document control
Owner: [POLICY_OWNER_ROLE, e.g. AI Officer]
Approved by: [APPROVER_NAME_AND_ROLE]
Version: [VERSION]
Effective date: [EFFECTIVE_DATE]
Next review: [NEXT_REVIEW_DATE]

AI Compliance Starter Kit. Template for the impact assessment of an AI system. It covers the fundamental rights impact assessment under Art. 27 of Regulation (EU) 2024/1689 where it applies as a legal duty, and the impact assessment under ISO/IEC 42001 A.5 for all other systems. Sections 2 to 7 follow Art. 27(1)(a) to (f). One assessment is kept per system; a repeated assessment creates a new version.

1. System and occasion

Item Entry
AI system [SYSTEM]
Identifier in the AI register [IDENTIFIER]
Role of the organisation [ROLE, e.g. deployer]
Legal duty under Art. 27 [YES_OR_NO]
Reason for the legal duty [e.g. body governed by public law, public service, Annex III point 5(b) or (c)]
First use on [DATE]
Builds on [e.g. impact assessment of the provider or of a comparable system, otherwise "own assessment"]

A fundamental rights impact assessment is a legal duty for deployers that are bodies governed by public law or private entities providing public services when they deploy a high-risk AI system under Art. 6(2) (except Annex III point 2), and for deployers of systems under Annex III point 5(b) and (c). The assessment is carried out before the first use and updated as soon as any of the elements changes (Art. 27(2)).

2. Processes of use (Art. 27(1)(a))

[Description of the processes in which the system is used in line with its intended purpose: which workflow, which decisions it prepares or supports, who works with the results.]

3. Period and frequency (Art. 27(1)(b))

[Period of time within which the system is intended to be used, and frequency of use, e.g. "from [DATE], open-ended, around [NUMBER] cases per month".]

4. Affected groups of persons and impact areas (Art. 27(1)(c))

For each group, record whether it is vulnerable, which impact areas are affected and which benefit is expected for it. Vulnerable groups include in particular employees, children and minors, people with disabilities and older people.

Group Vulnerable Impact areas affected Expected benefit
[e.g. Customers] [YES_OR_NO] [selection from the list below] [BENEFIT]
[e.g. Older people] Yes [selection from the list below] [BENEFIT]

Groups to choose from: employees; applicants; customers; suppliers and partners; visitors; children and minors; people with disabilities; older people; general public.

Impact areas to choose from: fairness; accountability; transparency and explainability; security and privacy; safety and health; financial consequences; accessibility; human rights.

5. Risks of harm (Art. 27(1)(d))

Risk of harm Group affected Circumstances that may lead to risks Impact Likelihood
[e.g. unjustified disadvantage] [GROUP] [e.g. incomplete input data] [1 TO 5] [1 TO 5]

The information provided by the provider under Art. 13 is taken into account. Every impact above the acceptance threshold is entered as a risk scenario in the AI risk register.

Reasonably foreseeable misuse: [How the system could be used contrary to its intended purpose, and the risk scenario assigned in the AI risk register.]

6. Societal impact (ISO/IEC 42001 A.5.5)

Societal field Affected Description
Environmental sustainability [YES_OR_NO] [DESCRIPTION]
Economic effects [YES_OR_NO] [DESCRIPTION]
Government and public administration [YES_OR_NO] [DESCRIPTION]
Health and safety [YES_OR_NO] [DESCRIPTION]
Norms, culture and values [YES_OR_NO] [DESCRIPTION]

7. Human oversight (Art. 27(1)(e))

[Implementation of the human oversight measures according to the instructions for use: the person exercising oversight, their competence and authority, how they review results, override them and stop the system.]

8. Measures if the risks materialise (Art. 27(1)(f))

  • Mitigating measures: [measures that reduce the risks from section 5, with reference to the AI risk treatment plan]
  • Internal governance: [who decides when a risk materialises; escalation to top management]
  • Complaint mechanism: [how affected persons lodge a complaint and request an explanation under Art. 86]

9. Data protection impact assessment (Art. 27(4))

[Reference to the data protection impact assessment of the processing the system feeds into, with its status. What is recorded there is not collected again; this assessment adds the fundamental rights perspective.]

10. Notification to the market surveillance authority (Art. 27(3))

Item Entry
Notification required [YES_OR_NO]
Reason, if not required [e.g. exemption under Art. 46(1) or reliance on an existing assessment under Art. 27(2)]
Notified on [DATE]
Reference of the notification [REFERENCE]

The notification goes to the market surveillance authority that is also responsible for reporting serious incidents, using the completed template of the AI Office (Art. 27(5)).

11. Result and approval

Item Entry
Result [e.g. use approved, use after the measures are implemented, use discontinued]
Carried out by [NAME_AND_ROLE]
Approved by [NAME_AND_ROLE, top management for impacts above the acceptance threshold]
Completed on [DATE]
Next review [DATE, at the latest after one year]

The assessment is retained as a record per system for [RETENTION_PERIOD, e.g. 10 years].

Sources

Frequently asked questions

Who has to carry out a fundamental rights impact assessment?

Deployers that are bodies governed by public law or private entities providing public services when they deploy a high-risk AI system under Art. 6(2) (except Annex III point 2), and deployers of systems for creditworthiness evaluation and for risk assessment in life and health insurance (Annex III point 5(b) and (c)).

Do I need the template if there is no legal duty?

ISO/IEC 42001 requires in A.5 an impact assessment for every AI system, regardless of its risk class. The template covers both cases; without a legal duty, the notification to the market surveillance authority is omitted.

Can I reuse the data protection impact assessment?

Yes. Art. 27(4) provides that the fundamental rights impact assessment complements an existing data protection impact assessment. Refer to its sections and add the fundamental rights perspective, for example fairness, accessibility and societal impact.

No Cookies!

This wiki collects nothing, bakes nothing and leaves nothing behind. There's nothing to consent to. Privacy doesn't get better than this.