On 16 November 2026, the European Commission opened a consultation on the draft questionnaire and explanatory notice for fundamental-rights impact assessments under Article 27 of the EU AI Act.

The consultation is a proposal-stage development. It does not create a new duty and the draft template is not yet the final instrument. The underlying Article 27 obligation is part of binding EU law, however, and the Commission says affected deployers will need to comply from 2 December 2027.

For organisations likely to deploy in-scope high-risk systems, the draft is useful now. It shows the evidence, internal ownership and provider information that a defensible assessment is likely to require before deployment.

What the Commission has published

The AI Office has developed a draft Article 27 questionnaire with input from the EU Agency for Fundamental Rights. A draft explanatory notice addresses who must assess, when the assessment is required and how the deployer's duty relates to information supplied by the AI-system provider.

The Commission is seeking public feedback until 15 January 2027. It has also invited deployers that expect to fall within Article 27 to test the draft on one of their own systems. Expressions of interest for testing close on 4 December 2026.

Participation in the testing exercise is voluntary and does not alter the legal obligations. Once adopted, the final template is expected to be available as an automated tool on the AI Act Single Information Platform.

Who Article 27 covers

Article 27 is not a general assessment duty for every AI user. It applies before deployment of specified Annex III high-risk systems by bodies governed by public law, private entities providing public services, and deployers of the systems identified in Annex III points 5(b) and 5(c). Those latter categories concern creditworthiness or credit scoring, subject to the Act's stated exception, and risk assessment and pricing in relation to life and health insurance.

The provision generally concerns systems classified as high-risk under Article 6(2). Article 27 expressly excludes the Annex III point 2 critical-infrastructure category from this particular assessment route. Separate obligations under the AI Act and other laws may still apply.

A business should therefore determine its role, system purpose, Annex III category and deployment context before assuming that a fundamental-rights impact assessment is mandatory. Commercial labels such as platform, tool or assistant do not decide the legal classification.

What the binding assessment must address

Article 27 requires the assessment before the high-risk system is deployed. Its required content includes the deployer's process and intended purpose, the duration and frequency of use, affected people and groups, specific risks of harm, human-oversight measures and the measures to be taken if those risks materialise.

The deployer must notify the market-surveillance authority of the assessment results, subject to the detailed terms of Article 27. Information must be updated if relevant elements change or cease to be current during use.

A data protection impact assessment and an Article 27 assessment are related but not interchangeable. Where an Article 27 requirement has already been addressed through a DPIA under the GDPR or the Law Enforcement Directive, the AI Act permits cross-references or inclusion of the relevant material. The fundamental-rights analysis can therefore build on privacy governance, but its scope is wider than personal-data risk alone.

Draft guidance is not final law

The consultation document and questionnaire remain drafts. Stakeholder feedback may change wording, examples and the practical workflow before adoption. Organisations should not describe a completed draft questionnaire as regulatory approval, certification or proof of compliance.

The binding legal proposition comes from Article 27 and the applicable provisions of the AI Act. The Commission's draft is a proposed compliance aid produced under Article 27(5). Where there is tension between a template and the legislation, the legislation prevails.

It is nevertheless reasonable to use the draft for readiness work. That is an inference from the Commission's stated purpose and the structure of the questionnaire, not a legal requirement to adopt the draft before it is final.

Practical actions for AI founders and technology businesses

  1. Map each intended deployment against Article 6, Annex III and Article 27, recording the provider and deployer roles separately.
  2. Identify whether the customer is a public-law body, a private provider of public services, or a deployer of an in-scope credit or life and health insurance system.
  3. Give customers usable Article 13 information on intended purpose, limitations, foreseeable risks, input requirements, performance and human oversight.
  4. Design procurement and implementation timelines so the assessment is completed before deployment, not after operational launch.
  5. Document affected groups, including people who are not direct users but may experience the system's decisions or downstream effects.
  6. Connect the assessment to testing, accessibility, non-discrimination, complaints, human review, incident response and remediation controls.
  7. Align the FRIA with any GDPR data protection impact assessment while preserving analysis of non-privacy fundamental rights.
  8. Define change triggers for reassessment, including a new purpose, user group, model, data source, workflow, geography or provider update.
  9. Review supplier contracts for access to the evidence needed by deployers and for cooperation when risks, incidents or material changes arise.
  10. Consider responding by 15 January 2027, or applying by 4 December 2026 to test the draft, if the organisation expects to be directly affected.

The compliance conclusion

The Commission's consultation does not accelerate the Article 27 start date or expand the statutory scope. Its immediate significance is operational: affected deployers can now test whether their governance files contain the information needed for a structured fundamental-rights assessment.

Providers are not automatically responsible for completing their customers' assessments, but their documentation and contractual cooperation will often determine whether deployers can do the work properly. Founders selling into sensitive public-service, credit or insurance settings should treat FRIA readiness as a product and procurement issue, not only a legal formality.

Legal position and sources reviewed on 17 November 2026. This article provides general information only and does not constitute legal advice on any particular AI system, deployment or organisation.

This article provides general information and does not constitute legal advice on a specific matter. Legal obligations depend on the facts and the applicable law.