In this article9 sections
Two Assessments, One System, Real Confusion
The dpia vs fria eu ai act question is one of the most common ones HR Directors and Compliance Officers ask us. It is easy to see why. You deploy an AI tool for recruitment screening or employee performance monitoring. Someone flags that you need a DPIA under GDPR. Someone else says you also need a Fundamental Rights Impact Assessment under the EU AI Act. You wonder if these are the same document with different names, or two entirely separate exercises. They are neither. They overlap, they are not identical, and most private companies only need the first one.
This post explains the legal basis for each, who actually needs an FRIA, and how to build one hybrid document if you do need both.
What Is a DPIA and When Is It Required?
A Data Protection Impact Assessment is required under Article 35 of the GDPR whenever processing is likely to result in a high risk to the rights and freedoms of natural persons. Systematic and extensive evaluation of individuals based on automated processing, large-scale processing of sensitive data, and systematic monitoring of public areas all trigger this obligation.
In practice, if you use AI to screen CVs, score employee performance, or analyse communication patterns, a DPIA is almost certainly mandatory, and it is mandatory today. Your Data Protection Authority, such as the Dutch Autoriteit Persoonsgegevens or the Belgian APD/GBA, supervises this under the GDPR.
The DPIA asks: what are the risks to personal data, and how do you mitigate them? It looks at data flows, retention periods, lawful basis, data subject rights, and proportionality. It is privacy-centric.
What Is an FRIA and When Is It Required?
A Fundamental Rights Impact Assessment is an obligation introduced by the EU AI Act. Article 27 of the AI Act requires certain deployers of high-risk AI systems listed in Annex III to carry out an FRIA before putting those systems into use.
Not every deployer. Article 27(1) limits the duty to:
-
bodies governed by public law, such as municipalities and government agencies;
-
private entities providing public services, for example in education, healthcare or social housing;
-
deployers of AI used to evaluate the creditworthiness or establish the credit score of natural persons (Annex III, point 5(b)), or for risk assessment and pricing in life and health insurance (point 5(c)).
Critical infrastructure systems (Annex III, point 2) are excluded. A private employer using recruitment, task allocation or performance-monitoring AI does not fall into any of these groups. That AI is high-risk under Annex III, point 4, and the employer has deployer duties under Article 26, but no FRIA duty.
The FRIA asks: what are the risks to fundamental rights broadly, including non-discrimination, dignity, and fairness? It is wider than the DPIA. It covers risks to people who may not even be data subjects in the traditional sense.
Timing matters too. Article 27 sits in Chapter III of the Act. Since the Digital Omnibus on AI (Regulation (EU) 2026/1744), Chapter III applies to Annex III systems from 2 December 2027. High-risk systems already placed on the market or put into service before that date are only covered if they are significantly changed afterwards (Article 111(2)); systems intended for use by public authorities must comply by 2 August 2030 in any case.
Side-by-Side: DPIA vs FRIA
The key insight: the DPIA and FRIA share common ground but are not substitutes for each other. If you are in the Article 27 group, you will usually need both, and the AI Act lets you reuse your DPIA work.
The Article 27(4) Overlap Provision
This is the good news. Article 27(4) of the AI Act, as amended by Regulation (EU) 2026/1744, says that if any FRIA obligation is already met through a DPIA under Article 35 of the GDPR, the deployer may include cross-references to the relevant sections of that DPIA, or include relevant parts of it, in the FRIA. You are not starting from scratch. Article 27(5) adds that the AI Office will develop a questionnaire template, including an automated tool, to help deployers complete the FRIA in a simplified manner.
Elements you can typically carry over from your DPIA:
-
Description of the AI system and its purpose (identical in both)
-
Categories of individuals affected (data subjects map to rights-holders)
-
Data flows and processing logic (DPIA detail feeds FRIA system description)
-
Risk identification methodology (same structured approach)
-
Mitigation measures already identified (data minimisation, access controls, human oversight)
What the FRIA adds that the DPIA does not cover:
-
Assessment of non-discrimination and equality risks beyond personal data
-
Risks to freedom of expression, assembly, or access to justice
-
Impact on vulnerable groups who may not be direct users or data subjects
-
A description of human oversight measures and of what you will do if risks materialise, including internal complaint handling
In short, your DPIA is a strong foundation. The FRIA builds one additional floor on top of it.
When You Need Both
You need a DPIA and an FRIA (separate or integrated) when all of the following are true:
-
Your AI system processes personal data in a way that is likely to result in high risk under the GDPR.
-
The system is listed in Annex III of the AI Act as high-risk.
-
You are a deployer in the Article 27 group: a public body, a private provider of public services, or a deployer of credit-scoring or life and health insurance pricing AI.
Common scenarios:
-
A municipality using AI to assess benefit applications. Public body, Annex III essential public services. Both assessments required, the FRIA from 2 December 2027.
-
A lender scoring consumers' creditworthiness. Annex III, point 5(b). Both required, the FRIA from 2 December 2027.
-
A private company using recruitment AI that scores candidates based on CVs. High-risk under Annex III, point 4(a). DPIA almost certainly required. No FRIA duty, though the fundamental-rights questions are worth covering inside your DPIA.
If your AI system is not high-risk, or you are not in the Article 27 group, you need at most the DPIA.
A Practical Hybrid Template Approach
If you do need both, the most efficient approach is a single combined document with clearly labelled sections. Structure it like this:
Section 1: System description (shared by both DPIA and FRIA) Name the AI system, its provider, its purpose, the processes it is used in, how often, the categories of people it affects, and how decisions are made or supported.
Section 2: Data protection assessment (DPIA core, GDPR Art 35) Lawful basis, data minimisation, retention, data subject rights, security measures, DPO consultation.
Section 3: Fundamental rights assessment (FRIA core, AI Act Art 27) Specific risks of harm to the groups affected, beyond privacy. Discrimination analysis. Human oversight measures. What you do if risks materialise, including complaint handling. Cross-reference Section 2 where it already answers a question.
Section 4: Risk register and mitigations (shared) A single risk register referencing both frameworks. Each risk flagged by which instrument requires it to be addressed.
Section 5: Review schedule Review both when the system or its use changes materially. A fixed annual review is good practice.
This format uses the Article 27(4) option to cross-reference the DPIA. It also gives your data protection authority and market surveillance authority a single coherent document if they ask.
One Action to Take This Week
Pull up the list of AI systems your company currently uses. For each one, answer three questions: does it process personal data in a high-risk way, is it in Annex III of the AI Act, and are you in the Article 27 group? The DPIA question needs an answer now. The FRIA question needs one before 2 December 2027, but the inventory is the same work, so start it now.
For a free 2-minute check on where your AI tools stand, use the compliance tool at comply.khairos.ai. It asks plain-language questions and gives you a clear answer with no jargon.
References
Not sure where your company stands?
Answer 15 questions about how your team uses AI. You get a compliance score and the gaps to close first, as a PDF.