The European Commission published a news article on 16 November 2026 about a draft template for the fundamental rights impact assessment under Article 27 of the AI Act. It collects feedback on a questionnaire that helps certain deployers of high-risk AI systems prepare that assessment. The template is Commission guidance, and no standards body issued it. It still raises the question every standard raises: it tells an organization what to answer and leaves open who inside the organization decides what the answers mean. That gap matters because, once the template is final, the results go to the market surveillance authority, and someone inside the deployer has to stand behind them.
This column read the notice page in full and relies on it alone. Nothing below describes the draft template or the explanatory notice beyond what the notice says about them.
What the notice says
The draft template was developed by the AI Office under Article 27(5) of the AI Act with input from the EU Agency for Fundamental Rights. It follows the structure of Article 27(1) and sets out questions to guide deployers through the assessment. A draft explanatory notice accompanies it and gives guidance on which deployers need to conduct the assessment, when to conduct it, and how the Article 27 obligation relates to the obligations of the provider of the high-risk AI system.
On timing, the assessment has to be done before deployment of the system, and the deployer updates the information in the template if it considers that a listed element has changed or is no longer up to date. Deployers will need to comply with this obligation as of 2 December 2027. Once the template is finalised, deployers will need to use it for the fundamental rights impact assessment and notify the results to the market surveillance authority under Article 27(3).
The notice also describes a tool. Once adopted, the template will be set up as an automated tool on the AI Act Single Information Platform that will allow deployers to fill in the template easily.
The notice names the deployer and leaves the decider open
In the notice, the party that carries out the assessment is the deployer. Read in full, the notice page does not name who inside the deployer answers the questions or who accepts, modifies, or rejects the result. That is a reading of the notice page alone, and the draft documents may say more.
A form tool is machine support. Software that checks the completed form with no human review is Responsible AI, machine checking machine. Whether the answers are good enough, whether a gap in the evidence is acceptable, and whether a deployment should go ahead are human calls. This column’s recommendation, under AI Governance, is that one named person makes each of those calls, records the reason, and answers for it. The Commission’s notice does not require that. Basil Puglisi’s public technical comment on how ISO/DIS 21520 defines artificial intelligence governance rests on the same point: a definition or a template can list the activities, and an organization still has to say who decides.
How to comment and test
The draft template is open for public feedback until 15 January 2027. Comments go through the survey form linked from the Commission’s notice page.
The AI Office is also looking for deployers that will have to carry out the assessment and want to test the draft template in practice. Interested deployers can express interest by 4 December 2026. Participants apply the template to one of their own systems, and the AI Office supports them with an introductory session, replies to their questions, and a closing session to discuss the results. Depending on the number of expressions of interest, the AI Office may have to select the cases for testing. Participation is voluntary and without prejudice to obligations under the AI Act.
What to put in place now
This is not legal advice. The template is a draft, and counsel should confirm which Article 27 duties apply to your organization. Taking part in the testing does not change obligations under the AI Act, according to the notice.
Whatever the final template asks, an organization can set up the human side before December 2027:
- Name one owner for each system that will be assessed, with a deputy. The owner decides accept, modify, or reject on the assessment result before deployment.
- Record each decision with the evidence behind the answers and the reason for the decision. Keep the external sources the answers relied on in a retrievable form. The source-custody method Basil Puglisi calls HAIA-SCOPE is one way to do that.
- When the template’s questions cannot be answered as asked, or an internal control departs from what the final template and its explanatory notice say, write a signed exception with a review date. Send the same difficulty to the AI Office through the feedback survey before 15 January 2027.
- Re-open the decision when the system, its use, or the affected people change, since the notice says the information must be updated when an element changes.
Three measures show whether the owner is real:
- Share of systems in scope with a named owner and deputy.
- Share of failed or borderline answers with a recorded accept, modify, or reject.
- Days from a failed or borderline answer to a signed decision.
If the first number is below 100 percent, some assessments have no one on record who decided.
Sources
- European Commission. (2026, November 16). Commission seeks feedback on draft template for the fundamental rights impact assessment under the AI Act. Shaping Europe’s digital future. https://digital-strategy.ec.europa.eu/en/news/commission-seeks-feedback-draft-template-fundamental-rights-impact-assessment-under-ai-act
- Puglisi, B. C. (2026, September 15). ISO/DIS 21520 comment: Defining AI governance. basilpuglisi.com. https://basilpuglisi.com/iso-21520-ai-governance-comment/
- Puglisi, B. C. (2026, June 2). SCOPE: Source custody observable publication evidence. basilpuglisi.com. https://basilpuglisi.com/scope-source-custody-observable-publication-evidence/
FAQ
What did the Commission publish on 16 November 2026?
A news article collecting feedback on a draft template for the fundamental rights impact assessment under Article 27 of the AI Act. The template follows the structure of Article 27(1) and sets out questions to guide deployers through the assessment. An open call for testing runs alongside the consultation.
When can I comment, and when must I apply to test the draft?
Public feedback is open until 15 January 2027. Deployers interested in testing the draft template can express interest to the AI Office by 4 December 2026. The AI Office may select cases for testing depending on the number of expressions of interest.
When do deployers have to comply with the Article 27 obligation?
The notice states that deployers will need to comply as of 2 December 2027. The assessment has to be done before deployment of the system, and the information is updated if a listed element changes or is no longer up to date.
Will the final template be a software tool?
The notice says that once adopted, the template will also be set up as an automated tool on the AI Act Single Information Platform that will allow deployers to fill in the template easily. The notice page does not say who inside a deployer should use the tool or approve what is entered.
Does testing the draft protect an organization from liability?
The notice says participation is voluntary and without prejudice to obligations under the AI Act. Nothing in it gives a safe harbor. This is not legal advice, and counsel should confirm what applies.
Who should own the assessment inside an organization?
The notice page does not say. A practical approach is one named owner per assessed system, with a deputy, who decides on the result and records the evidence and the reason. The open question for the final template and the testing round is whether they will say who inside a deployer stands behind the answers, and deployers can raise it through the feedback survey.
AI disclosure: This column was researched and drafted with AI assistance by HAIA Agents (RAI by GrokBot), using primary sources verified at time of publication. #AIgenerated
Leave a Reply