Regulation (EU) 2024/1689 · Artificial Intelligence Act ● Annex III: — days

EU AI Act Complaint Process: How Article 85 Complaints Work

Individuals can challenge AI decisions in the EU: Article 85 lets them complain to authorities, Article 86 gets an explanation. Here is how the process works.

Renomira Team · 6 min read
Complaint form and pen on a desk
Photo: Smuconlaw / CC BY-SA 4.0 via Wikimedia Commons

The eu ai act complaint process is one of the least understood parts of the Regulation: people know the fines, few know that individuals can walk into a national authority and complain about an AI system directly. Article 85 gives every individual the article 85 ai act right to lodge a complaint with the relevant market surveillance authority, and Article 86 adds a right to explanation of individual decision-making. This guide covers who can complain and about what, what happens after a complaint lands, how can i challenge an ai decision eu questions actually resolve across the overlapping regimes, and what a company should do when the letter arrives.

Article 85: the right to complain

The provision is deliberately broad. Any individual who considers that the Regulation has been infringed may lodge a complaint with the market surveillance authority of the member state where the alleged infringement occurred — or where the affected person resides or works, where the Regulation so provides. Key features:

  • Who: any individual — no consumer status, no data-subject status, no loss threshold required. Trade associations and NGOs pursue parallel routes (the Regulation also protects the representative actions of qualified entities under the broader enforcement architecture).
  • About what: any suspected infringement of the Regulation — a prohibited practice operating nearby, an unmarked high-risk system making decisions about the complainant, missing transparency notice, refusal to explain.
  • With whom: the national market surveillance authority competent for the system’s provider or deployer in that member state.
  • Consequences for the individual: none for complaining. There is no fee, no risk of adverse cost, and retaliation concerns fall under existing employment and GDPR protections.

What the authority does with a complaint is discretionary: it must handle complaints in accordance with its established procedures, examining the facts and providing the complainant with information about progress. Complaints are an input to its supervisory priorities — but a well-documented complaint about a high-risk system in a regulated domain is precisely the trigger that opens a larger inquiry.

Article 86: the right to explanation of individual decision-making

Complaint is the reactive tool; Article 86 is the proactive one. Subject to conditions and exceptions laid down in the Regulation, affected persons may obtain clear and meaningful explanations of the role of the AI system in the decision-making procedure from the deployer, when that system is used to take a decision in their case about a high-risk system’s output producing legal effects or similarly significantly affecting them.

Practical boundaries worth knowing before you ask:

  • It applies to deployers of high-risk systems making decisions about individuals — not to every chatbot or recommendation engine.
  • Exemptions exist where the explanation would jeopardise the system’s effectiveness or reveal trade secrets, and where the decision is authorised by law.
  • The right sits alongside GDPR Article 22 (logic involved, automated decision-making safeguards) — often the stronger instrument where solely automated decisions with legal or similarly significant effects are in play.
  • Ask in writing, keep the request, and set expectations: a deployer that has documented its oversight and feature logic can answer; one that never inspected the model will struggle — and that struggle is itself a finding.

How the routes compare

Route Against what Who you approach What you can get
Article 85 complaint Suspected infringement of the AI Act National market surveillance authority Investigation, information on progress, possible corrective action
Article 86 explanation Role of a high-risk AI system in a decision about you The deployer Clear, meaningful explanation of the system’s role
GDPR access/complaint Personal data processing and automated decisions Controller, then DPA / courts Access, logic description, human review, compensation
Court action Damages from non-compliant AI Civil courts (with AI Liability Directive developments reshaping evidence rules) Remedies and damages

Sophisticated complainants stack them: an Article 86 request to the deployer to pin the factual record, a GDPR access request to get the data and the logic, then an Article 85 complaint carrying that evidence to the authority. Each step creates documents the others rely on.

What happens after a complaint is filed

The process a complainant should expect, in order:

  1. Receipt and routing. The authority acknowledges and checks competence — a complaint about a system operated in another member state may be forwarded; follow it.
  2. Examination. The authority examines the facts and may come back for more: the decision notice, the vendor’s name, dates, screenshots, correspondence with the deployer. The more of this you attach up front, the faster the file moves.
  3. Informal resolution. Some authorities invite the operator to respond or correct — a transparency notice added, a process fixed — before formal steps.
  4. Formal supervisory measures where warranted: information demands, a commitment to comply, corrective action, restriction or prohibition of the system, and in serious cases fines under Article 99 (prohibited practices at EUR 35,000,000/7%, transparency breaches at EUR 15,000,000/3%).
  5. Information back to you. The authority provides information about the progress of handling — you are not left in silence, though “progress” is their definition of it.

Complaints also feed priority-setting: patterns of similar complaints about the same tool become the basis for a sweep. That is the quiet strategic value — an individual complaint is rarely the whole case, but it can be the first documented one.

What companies should do when one arrives

Receiving an Article 85 complaint is not a crisis — mishandling it is. The protocol: treat it as a regulator communication (preserve relevant documents immediately; suspend routine deletion for the systems named), route it to the accountable owner rather than an inbox, and assemble the facts before responding — classification record, oversight assignment, transparency notices shown, logs. Check the complaint against your register: if the system is not on it, fix the register first internally and decide how to explain the discrepancy honestly. Reply within the stated window with the direct answer first, then evidence. If the complaint reveals a real gap, open the corrective action the same week and record it — documented self-correction mitigates outcomes. Finally, log the complaint itself in your governance minutes: authorities ask whether operators learn, and a complaint that produced a dated process change is a strong answer.

Where complaints go wrong (and how to strengthen yours)

Complaints fail for avoidable reasons: aimed at the wrong body (a court, a company, the wrong member state), describing a bad outcome without identifying the Regulation provision allegedly infringed, or attaching no evidence at all. Strengthen yours with four moves. First, identify the system and operator — company name, product name, the decision you received; if the notice did not say AI was involved, say so, because opacity is itself a transparency question under Article 50. Second, cite the provision: Article 85 covers any infringement, but naming the suspected one (transparency under Article 50, deployer duties under Article 26, a prohibited practice under Article 5) focuses the file. Third, build the timeline: decision date, request for explanation, refusal or silence, complaint. Fourth, route correctly — the authority competent where the operator is established or the decision affected you; a complaint sent to the wrong member state is not a complaint anyone must chase on your behalf. Keep a copy of everything you send: the sequence of documents you create becomes the record any later route — court, data protection authority — will use.

How to check your own position

Run the complainant’s test on your own systems: pick the three that make or inform decisions about individuals, and write what a determined person could learn in ten minutes — is the AI involvement disclosed, is there an explanation route, would your oversight records show how the decision was reached? Then send yourself the Article 86 letter: draft the response you would have to give, and notice which parts you cannot answer from documentation. If your deployer team would have to improvise an explanation of a high-risk system’s role, that improvisation is the gap — close it with a written decision-logic summary per system. Finally, confirm your complaints handler exists: one named person, a preservation rule, and a response template. Firms that practise receiving complaints respond in days; firms that improvise respond late, and lateness is what enforcement remembers.

Frequently asked questions

How do I complain about an AI decision under the EU AI Act?

Under Article 85, lodge a written complaint with the market surveillance authority of the member state where the infringement occurred or where you reside or work. Describe the decision, the system involved, the deployer or provider if known, and attach evidence — notices, letters, screenshots, dates. There is no fee, and the authority must handle the complaint under its procedures and inform you of progress.

What is the difference between Article 85 and Article 86?

Article 85 is a complaint to a public authority about a suspected infringement of the Regulation by anyone — it triggers supervisory action. Article 86 is a private right against the deployer: when a high-risk AI system played a role in a decision about you with legal or similarly significant effects, you can request a clear, meaningful explanation of that role — subject to trade-secret and legal-authority exceptions.

Can an EU AI Act complaint get my decision reversed?

The AI Act’s complaint route delivers investigation and enforcement, not an order to reverse your individual decision. For reversal or human review, pair it with GDPR Article 22 where a solely automated decision with legal or similarly significant effects is involved. In practice the combination works: the explanation request and access request build the record, the complaint pressures the operator, and the data protection route delivers individual remedies.

Does the AI Act require companies to explain automated decisions?

Partly. Article 86 gives affected persons the right to clear and meaningful explanations of the role of a high-risk AI system in a decision about them, when deployed for decisions producing legal or similarly significant effects — with exceptions for trade secrets and legal authorisation. The GDPR’s automated-decision provisions go further on the logic involved and human intervention, so most EU complainants hold both levers at once.

What penalties apply when a company ignores a complaint-driven finding?

It depends on the underlying infringement: prohibited practices reach EUR 35,000,000 or 7% of worldwide annual turnover, transparency and most operator-duty breaches EUR 15,000,000 or 3%, and supplying incorrect or misleading information to authorities its own tier under Article 99. The aggravating factor is typically obstruction — cooperation and prompt corrective action are visible to the authority and reflected in the assessment of fines.

Next step

Put this guide to work on your own estate.

Run the free exposure scan to see which of these articles actually apply to your company — 12 questions, four minutes, no card. Confirm your email and the full report unlocks.

  • Free forever plan
  • No credit card
  • EU-hosted · GDPR-native
Run a free scan Sign up free

Build a live AI register in minutes — upgrade only when you need the audit pack.

eu ai act complaint processarticle 85 ai act right to lodge a complainthow to complain about an ai decision europecan i challenge an ai decision eumarket surveillance authority complaint ai actlodging a complaint against an ai systemindividual rights under the eu ai actai act right to explanation article 86

Keep reading