A comprehensive guide for businesses to understanding Article 22, the prohibition on solely automated decisions, the available exceptions, required safeguards, and the separate right to object under Article 21.
1. Introduction: Why Automated Decisions Matter
Across every sector of the modern economy, organisations are deploying algorithms, machine learning models, and automated scoring systems to make or inform decisions that directly affect individuals. Credit institutions use algorithmic models to approve or decline loan applications. Insurers calculate personalised premiums through actuarial software with minimal human involvement. Employers filter job applications using AI-powered screening tools. Online platforms rank and serve content based on inferred behavioural profiles. In each of these scenarios, decisions with significant real-world consequences are being made at scale, at speed, and with little or no human deliberation.
The General Data Protection Regulation (GDPR) responds to this reality with two distinct but complementary protections. Article 22 establishes a qualified right not to be subject to a decision based solely on automated processing — including profiling — where that decision produces legal effects or similarly significant effects on the individual. Article 21 provides a separate and broader right to object to processing carried out on certain legal bases, including processing for the purposes of profiling. Together, these provisions represent some of the most consequential and technically demanding obligations in data protection law, and compliance failures have attracted significant regulatory attention.
2. Profiling and Automated Processing: Key Definitions
Before examining Article 22 itself, it is important to establish what is meant by the key terms. Article 4(4) GDPR defines “profiling” as any form of automated processing of personal data consisting of the use of personal data to evaluate certain personal aspects relating to a natural person — in particular to analyse or predict aspects concerning that person’s performance at work, economic situation, health, personal preferences, interests, reliability, behaviour, location, or movements. The definition is intentionally broad. Any automated evaluation that categorises, scores, or draws inferences about an individual on the basis of their personal data will constitute profiling, even where the categories or predictions used may appear neutral on their face.
It is important to understand that profiling and automated decision-making are not synonymous, and Article 22 does not apply to all profiling. A business may lawfully profile individuals — gathering and analysing personal data to build up a picture of their characteristics or likely behaviours — provided it does so on an appropriate legal basis and complies with the transparency requirements set out elsewhere in the Regulation. Article 22 is engaged only at the point where such profiling (or any other automated processing) feeds into a decision that produces legal or similarly significant effects on the individual, and where that decision is made solely on the basis of automated processing with no meaningful human involvement.
3. The Scope of Article 22: When Does the Prohibition Apply?
Article 22(1) provides that a data subject shall have the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning them or similarly significantly affects them. Three cumulative conditions must be satisfied before Article 22 is engaged: the processing must be automated; the decision must be based solely on that automated processing; and the decision must produce legal effects or similarly significant effects.
The requirement that a decision be based “solely” on automated processing is one of the most contested and practically significant aspects of Article 22. The EDPB’s guidelines make clear that human involvement must be genuine, meaningful, and consequential — not merely nominal or perfunctory. A business cannot satisfy this requirement by having a human employee formally “approve” each decision without actually reviewing the substance of the automated output, scrutinising the data that informed it, or having the authority and capability to reach a different conclusion.
In practice, this means that many automated systems that are technically described as “decision support tools” may still fall within Article 22 if the human nominally involved in the process lacks the time, information, or authority to meaningfully override the algorithmic output. Businesses relying on a human review step as their primary means of avoiding Article 22’s application should carefully audit the reality of how that review operates.
4. What Constitutes a “Legal or Similarly Significant Effect”?
Even where processing is fully automated, Article 22 is only engaged if the resulting decision produces “legal effects” or “similarly significantly affects” the individual. The first category — legal effects — is relatively clear: decisions that affect the legal rights or legal status of an individual, such as denying a person access to a public benefit, terminating a contract, or triggering a regulatory obligation, will plainly produce legal effects.
The second category — “similarly significant effects” — is less precisely defined but has been interpreted broadly by regulators. The EDPB’s guidelines identify a number of factors that indicate a similarly significant effect: whether the decision has a substantial impact on the individual’s circumstances, behaviour, or choices; whether the decision is likely to have a prolonged or permanent effect; whether the decision results in discrimination or unequal treatment; and whether the individual is likely to have been aware that the decision would be made about them on the basis of automated processing.
Examples of decisions likely to produce similarly significant effects include the automatic refusal of an online credit application, the denial of employment through an automated recruitment tool, the setting of insurance premiums by algorithmic assessment, and the exclusion of a person from access to an online service or platform based on automated profiling.
5. The Three Exceptions to the Article 22 Prohibition
Article 22(1) establishes what the EDPB and most commentators have characterised as a prohibition in substance, but Article 22(2) provides three exceptions under which such processing may nonetheless be carried out.
Exception 1 (Art. 22(2)(a)) — Necessary for a Contract: The decision is necessary for entering into, or performance of, a contract between the data subject and a data controller. The necessity requirement must be genuine: the contract must not be reasonably achievable without the automated decision-making in question.
Exception 2 (Art. 22(2)(b)) — Authorised by Union or Member State Law: The decision is authorised by Union or Member State law to which the controller is subject, which also lays down suitable measures to safeguard the data subject’s rights and freedoms and legitimate interests.
Exception 3 (Art. 22(2)(c)) — Based on Explicit Consent: The decision is based on the data subject’s explicit consent. Consent must meet the full requirements of Article 7 GDPR and must be specific to the automated decision-making in question. It must be freely given, informed, and unambiguous.
Even where one of the three exceptions applies, the controller must still implement “suitable measures to safeguard the data subject’s rights and freedoms and legitimate interests” as required by Article 22(3). The exceptions do not provide a licence to conduct automated decision-making without further obligations — they merely remove the blanket prohibition.
6. Mandatory Safeguards Under Article 22(3)
Where a controller relies on the contractual necessity or explicit consent exceptions in Article 22(2)(a) and (c), Article 22(3) requires the controller to implement suitable measures to safeguard the data subject’s rights, freedoms, and legitimate interests. The Regulation specifies, as a minimum, three such measures: the right to obtain human intervention on the part of the controller; the right to express one’s point of view; and the right to contest the decision.
Human Intervention: The data subject must have access to a mechanism by which a human reviews the automated decision. The human reviewer must be capable of genuinely reassessing the decision based on the individual’s specific circumstances, not merely confirming or explaining the algorithmic output.
Expressing a Point of View: The data subject has the right to express their point of view before or about the automated decision. This requires the controller to provide a meaningful opportunity for the individual to submit information or representations that could influence the outcome.
Contesting the Decision: The right to contest the decision supplements the right to human review by providing a formal route for the data subject to challenge the automated determination. Controllers should ensure they have a documented process for handling such challenges.
7. Special Category Data and the Heightened Restrictions
Article 22(4) imposes a significantly stricter regime where the automated decision-making involves the processing of special categories of personal data — that is, data revealing racial or ethnic origin, political opinions, religious or philosophical beliefs, trade union membership, genetic data, biometric data processed for the purpose of uniquely identifying a natural person, health data, or data concerning a person’s sex life or sexual orientation. Where such data is involved, the contractual necessity exception in Article 22(2)(a) and the legal authorisation exception in Article 22(2)(b) are unavailable. Only explicit consent, or grounds of substantial public interest under Union or Member State law, may justify solely automated decisions based on special category data.
Data protection authorities have consistently taken the position that inferred data can constitute personal data — including special category data — even if it has not been directly provided by the individual. An automated system that infers health status, religious beliefs, or political opinions from other data will be processing special category data for Article 22(4) purposes, even if the input data is not itself special category.
8. The Right to Object Under Article 21
Separate from and complementary to the protections in Article 22 is the right to object under Article 21. This right operates at the level of the processing itself, rather than at the level of a specific automated decision, and applies more broadly to any processing carried out on the basis of legitimate interests (Article 6(1)(f)) or for the purposes of a task carried out in the public interest or in the exercise of official authority (Article 6(1)(e)).
The right to object to direct marketing — including profiling carried out to the extent it relates to direct marketing — is unconditional. Where a data subject objects to their data being used for direct marketing purposes, the controller must cease that processing immediately and without any assessment of whether the controller’s legitimate interests override the data subject’s objection.
For processing carried out on the basis of legitimate interests or public interest that does not relate to direct marketing, Article 21(1) provides a qualified right to object. Where a data subject objects, the controller must cease the processing unless it can demonstrate compelling legitimate grounds for the processing which override the interests, rights, and freedoms of the data subject, or the processing is for the establishment, exercise, or defence of legal claims.
Article 21(4) places a specific transparency obligation on controllers in respect of the right to object. The data subject must be explicitly brought to the attention of their right to object, in a manner that is clearly distinguishable from other information and presented at the latest at the time of the first communication with the data subject.
9. Transparency Obligations: Privacy Notices and the Right to Explanation
The transparency obligations in Articles 13 and 14 GDPR require controllers to provide specific information about automated decision-making in their privacy notices. Where a controller engages in automated decision-making — including profiling — that falls within Article 22(1) and (4), the controller must provide meaningful information about the logic involved, the significance of the processing, and the envisaged consequences of such processing for the data subject.
The phrase “meaningful information about the logic involved” requires more than a generic and uninformative description. The information provided must be sufficient for the data subject to understand in general terms what factors the automated system takes into account, how those factors are weighted or combined, and what the likely consequences are for the individual of various outputs.
10. Data Protection Impact Assessments and Automated Processing
Article 35(3)(a) GDPR provides that a Data Protection Impact Assessment (DPIA) is mandatory where a controller carries out systematic and extensive evaluation of personal aspects relating to natural persons which is based on automated processing, including profiling, and on which decisions are based that produce legal effects concerning the natural person or similarly significantly affect the natural person.
A DPIA in this context must systematically describe the envisaged processing operations and the purposes of the processing; assess the necessity and proportionality of the processing in relation to the purposes; assess the risks to the rights and freedoms of data subjects; and identify the measures envisaged to address the risks and demonstrate compliance with the Regulation.
11. Practical Steps for Businesses
- Conduct a decision-making audit: Map all automated systems and processes across the organisation that generate outputs affecting individuals. For each, assess whether the output constitutes a “decision” for Article 22 purposes, whether it is solely automated, and whether it produces legal or similarly significant effects.
- Identify and document the applicable exception: For any processing that falls within Article 22(1), identify and document the exception under Article 22(2) being relied upon.
- Implement and test the Article 22(3) safeguards: Establish documented processes for human review, the opportunity to express a point of view, and contesting the decision.
- Assess whether special category data is involved: Review automated systems for the processing of special category data, including inferred special category data.
- Update privacy notices: Ensure privacy notices include the information required by Articles 13(2)(f) and 14(2)(g), including a description of the logic involved in automated decision-making.
- Honour the right to object: Implement a process for receiving and acting on objections under Article 21.
- Carry out a DPIA: For automated decision-making within the scope of Article 22(1), carry out a DPIA as required by Article 35(3)(a).
- Review third-party arrangements: Where automated decision-making systems are operated by third parties, review contractual and data processing arrangements to ensure compliance obligations are appropriately allocated.
- Train relevant staff: Ensure that staff involved in automated decision-making processes receive adequate training on the requirements of Articles 21 and 22.
- Monitor and review: Build in regular review cycles to ensure that compliance assessments remain current and that safeguards continue to operate effectively.
12. Regulatory Guidance and Enforcement Trends
The EDPB has issued detailed guidance on automated decision-making and profiling under the GDPR, most notably in the Article 29 Working Party’s Guidelines on Automated individual decision-making and Profiling (WP251), which remain central to the interpretation of Articles 21 and 22. Businesses are strongly advised to consult this guidance when designing or auditing their automated processing activities.
Enforcement activity in this area has grown substantially since the GDPR came into force in May 2018. Significant enforcement action has been taken against organisations in the financial services, insurance, recruitment, and social media sectors. Regulators have found violations relating to inadequate transparency about automated processing, failure to provide genuine human review, reliance on consent that did not meet the required standard, and failure to carry out DPIAs for high-risk automated processing. Fines for serious violations of Article 22 can reach up to €20 million or four percent of total worldwide annual turnover — whichever is higher.
The EU AI Act, which entered into force in August 2024 and is being phased in on a rolling basis through to 2027, introduces a separate but overlapping regulatory framework for artificial intelligence systems, including those used in automated decision-making. Businesses subject to both the GDPR and the AI Act should undertake an integrated compliance review to identify synergies and gaps.
Legal Information Notice: This page has been prepared for general informational purposes only and does not constitute legal advice. The application of Articles 21 and 22 of the GDPR depends on the specific facts and circumstances of each organisation and each automated processing activity. Businesses should seek independent legal advice tailored to their particular situation.
