The AI Act provides several remedies that sit alongside market surveillance. Under Article 86, a person subject to a decision taken by a deployer on the basis of the output of a high-risk AI system listed in Annex III, other than a system in point 2, has a right to an explanation of the role of the system in the decision-making procedure, the main parameters of the decision and the related input data, subject to the provision's conditions and exceptions. Under Article 87, persons reporting infringements are protected under the whistleblower rules in Directive (EU) 2019/1937. Judicial remedies, including the right to an effective judicial remedy against a legally binding decision of a market surveillance authority, are preserved. These remedies are individual and procedural: they do not decide the underlying classification of a system.
Who has a right to an explanation?
Article 86 gives a right to an explanation to a person who considers that a decision taken by a deployer, which produces legal effects concerning them or similarly significantly affects them, was taken mainly on the basis of the output of a high-risk AI system listed in Annex III, other than a system listed in point 2. The right applies where the decision is based mainly on the system's output, not where the system played only a peripheral role, and it applies to decisions taken by deployers, not to every use of an AI system. The person must request the explanation, and the deployer must provide it, subject to the exceptions in the provision, including where Union or national law provides an exception or where the explanation would conflict with other legal obligations. A deployer should therefore be able to identify, for each consequential decision, whether a high-risk Annex III system was used, whether the decision was based mainly on its output, and what explanation can be given, before the first request arrives.
What must an explanation contain?
An explanation under Article 86 must cover the role of the AI system in the decision-making procedure, the main parameters of the decision and the related input data, in a clear and meaningful manner. It must explain how the system contributed to the decision, what the main factors were and what data about the person were used, without disclosing trade secrets or confidential information beyond what the provision requires. The explanation should be written for the person concerned, in language they can understand, and should allow them to assess whether the decision was lawful and whether to seek a remedy. A generic statement that AI was used, or a technical description of the model that does not connect to the decision, does not satisfy the provision. The deployer should keep a record of the explanation given, the decision concerned, the system and version used, and the date, so a later question about the explanation can be answered from the record rather than from memory.
For the broader regulatory framework, see enforcement and market surveillance.
What protection exists for persons reporting infringements?
Article 87 provides that Directive (EU) 2019/1937, the whistleblower directive, applies to the reporting of infringements of the Act and to the protection of persons reporting such infringements. That means a person who reports an infringement through an internal or external reporting channel, in accordance with the directive, is protected against retaliation, subject to the directive's conditions, and the organisation receiving the report must follow the directive's procedures for handling it. The protection covers reporting persons as defined in the directive, and it sits alongside the complaint route in Article 85, which is open to any person with grounds to consider an infringement occurred. An organisation should therefore maintain a reporting channel that complies with the directive, train the people who receive reports, and ensure that a report about an AI system is routed to the compliance owner promptly, with confidentiality protected and retaliation prohibited. A report that is ignored or mishandled can itself become a compliance issue.
What judicial remedies are available?
The Act preserves judicial remedies, including the right to an effective judicial remedy against a legally binding decision of a market surveillance authority, and the right to seek a remedy before a court where a person's rights under the Act are affected. The procedure, the court and the timescale depend on national law and on the nature of the decision or act challenged, and the Act's remedies sit alongside remedies under other law, including data protection, consumer, employment and fundamental rights law. A person considering a judicial remedy should seek qualified advice promptly, since time limits can be short and the correct forum depends on the decision and the Member State. An operator facing a judicial challenge should preserve the complete record for the decision concerned, including the classification, the system version, the logs, the explanation given and the authority correspondence, and should not destroy or alter records after proceedings are contemplated.
How should deployers prepare for explanation requests?
Deployers should prepare for explanation requests as part of deployment planning, not as an afterthought when a request arrives. For each high-risk Annex III use, the deployer should record the decision types affected, the system's role in each, the main parameters and input data categories, and the template or process for giving an explanation. Staff who take or communicate decisions should be trained to recognise an explanation request, to route it to the owner, and to respond within the timescale the deployer has set, with the explanation checked for accuracy and completeness before it is sent. The deployer should also coordinate with the provider, whose information about the system's capabilities, limitations and parameters may be needed for a complete explanation, while respecting confidentiality. A deployer that cannot explain a decision taken mainly on a system's output will struggle to show that its deployment was properly overseen, even if the system itself was compliant.
Continue with the right to lodge a complaint under Article 85.
What does this guide not decide?
This guide does not decide whether a particular person has a right to an explanation, what a particular explanation must say, or whether a particular report is protected under the whistleblower directive. Those questions depend on the decision, the system, the provision and the facts, and they require the consolidated text of Articles 86 and 87 and Directive (EU) 2019/1937 to be applied to the specific case. The guide also does not set timescales for explanations or judicial remedies, which depend on national law and the provision concerned. Where an explanation request, a report or a remedy is in issue, seek qualified advice promptly, and keep the complete record for the decision or report concerned.
Frequently asked questions
Does every AI-assisted decision carry a right to an explanation?
No. Article 86 applies where a deployer's decision producing legal or similarly significant effects was taken mainly on the basis of the output of a high-risk Annex III system, other than a point 2 system, subject to its conditions and exceptions.
Are whistleblowers protected when reporting AI Act infringements?
Yes. Article 87 applies Directive (EU) 2019/1937 to reporting infringements of the Act and to protecting persons who report them, subject to the directive's conditions.
Can a market surveillance decision be challenged in court?
Yes. The Act preserves the right to an effective judicial remedy against a legally binding decision of a market surveillance authority, under the procedure and timescale of national law.
Continue through this AI Act cluster
This article is an informational overview, not a legal determination. Check the current legal text and official implementation guidance for your system and role.
Consolidated Regulation (EU) 2024/1689, as amended ↗AI Act Service Desk implementation timeline ↗