Deployers who must carry out a fundamental rights impact assessment under the AI Act now have a draft of the form they will use. The European Commission published it on 16 November 2026, together with a draft explanatory notice. It is asking deployers to try it on one of their own systems.
That makes this the moment to find out where your own information runs out, more than a year before the obligation applies.
What the Commission published
The AI Office drafted the questionnaire template under Article 27(5) of the AI Act. The EU Agency for Fundamental Rights gave input. According to the Commission’s announcement, the template follows the structure of Article 27(1). Its questions guide the deployer through the fundamental rights impact assessment step by step. Once adopted, it will become an automated tool on the AI Act Single Information Platform, so deployers can complete it online.
The draft explanatory notice covers three questions. Which deployers have to carry out the assessment? When do they have to do it? And how does the deployer’s duty relate to the obligations of the provider of the high-risk system?
Two deadlines come with it. Deployers who want to test the draft on a real system can register their interest by 4 December 2026. Anyone can give feedback on the draft until 15 January 2027. The AI Office may select test cases to get a balanced mix of uses, deployer types, sizes and Member States. Taking part is voluntary and does not change anyone’s obligations.
Who needs a fundamental rights impact assessment
Article 27 does not apply to every deployer of a high-risk system. It covers bodies governed by public law and private entities that provide public services. It also covers deployers of two kinds of systems in Annex III. Point 5(b) is creditworthiness and credit scoring, with fraud detection excluded. Point 5(c) is risk assessment and pricing for life and health insurance. Systems used as safety components in critical infrastructure, Annex III point 2, are excluded.
So a bank that scores loan applicants with AI is in scope. A municipality using AI to decide on benefits is in scope. A private company that screens job applicants with AI deploys a high-risk system too. Yet it does not need a fundamental rights impact assessment under Article 27, unless it provides a public service.
The obligation applies from 2 December 2027. That date follows from the Digital Omnibus on AI, Regulation (EU) 2026/1744, which moved the Annex III rules from August 2026. We wrote about what that shift means for planning in Digital Omnibus: the AI Act delay that makes 2026 busier.
What a fundamental rights impact assessment contains
Article 27(1) lists six elements. The template will ask about each of them.
- A description of the deployer’s processes in which the system will be used, in line with its intended purpose.
- The period and frequency of intended use.
- The categories of people and groups likely to be affected.
- The specific risks of harm to them, using the provider’s information under Article 13.
- The human oversight measures, in line with the provider’s instructions for use.
- The measures to take if those risks materialise, including internal governance and complaint mechanisms.
The provider’s information is the starting point
Three of the six elements depend on what the provider supplies. They rely on the intended purpose, the Article 13 information and the instructions for use. If those documents are thin, the fundamental rights impact assessment will be thin too.
That is a procurement question before it is a compliance question. Ask the vendor for the instructions for use and the Article 13 information before you sign. Check that they describe your use case. If you change the purpose or modify the system, the provider obligations may move to you. We set that out in Your vendor can leave you holding the provider obligations. A vendor’s own code of conduct will not fill the gap. Our piece on Microsoft’s AI code of conduct shows how narrow such documents can be.
The DPIA does part of the work
Credit scoring, insurance pricing and public services all involve personal data. A data protection impact assessment under Article 35 GDPR may already exist. Article 27(4) says the fundamental rights impact assessment then complements the DPIA. Article 27(5) allows the template to refer to the relevant parts of it.
In practice, the two teams should work from one file. The privacy team knows the data flows and the people affected. The AI governance team knows the system and its oversight. Two separate documents mean describing the same process twice, with a risk that the versions disagree.
When the fundamental rights impact assessment is due
The duty to carry out a fundamental rights impact assessment applies to the first use of the system. In similar cases, a deployer may rely on an earlier assessment of its own. It may also rely on an existing assessment by the provider. When any of the six elements changes or goes out of date, the deployer updates the assessment.
After completing it, the deployer notifies the market surveillance authority of the results, using the template. Article 46(1) allows an exemption in exceptional cases.
The fifth element deserves more time than it usually gets. Human oversight has to describe what a person actually does with an output. A reviewer who only approves what the system proposes does not count. We made that point with the Uber decision in Meaningful human involvement is not a rubber stamp.
Why testing the template now pays off
Filling in a draft fundamental rights impact assessment for one real system shows quickly where the gaps are. The period and frequency of use and the groups affected are not in the vendor’s documents. Only the deployer can answer them.
Pick one system that is clearly in scope, such as a credit scoring tool. Collect the provider’s instructions for use, the Article 13 information and the DPIA if there is one. Add the names of the people who review the outputs. Then go through the six elements of the fundamental rights impact assessment and note what you cannot answer yet. Those gaps are your work plan for 2027 and the agenda for your next vendor review.
If you have a system in scope, you can register interest in testing the template until 4 December. You can give feedback on the draft until 15 January 2027.