---
title: Does GDPR Article 22 apply to decisions made by AI agents?
url: https://opencontroller.lyzr.com/blog/gdpr-article-22-and-ai-agents
published: 2026-11-15
updated: 2026-11-15
author: Lyzr
---

# Does GDPR Article 22 apply to decisions made by AI agents?

**GDPR Article 22 applies when an AI agent makes a decision about a person that is based solely on automated processing and produces legal effects or similarly significant effects, such as refusing credit, rejecting a job application or cancelling a contract. Such decisions are only allowed under one of three exceptions, and even then you must offer human intervention, a way to express a view and a way to contest the decision. Most agents that draft, summarise or recommend for a human reviewer fall outside Article 22, but only if the human review is real.**

This is general information, not legal advice. For the assessment that usually sits alongside an Article 22 analysis, see [DPIAs for AI agents](/blog/dpia-for-ai-agents), and for the wider framework view, [how to map AI agent controls to compliance frameworks](/blog/how-to-map-ai-agent-controls-to-compliance-frameworks).

## What does Article 22 actually say?

[Article 22(1)](https://gdpr-info.eu/art-22-gdpr/) gives a data subject 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.

Break that into three tests. All three must be met.

| Test | Question for your agent | Example that meets it |
|---|---|---|
| Decision | Does the agent's output settle an outcome for a person? | Agent approves or denies a refund claim |
| Solely automated | Is there no meaningful human involvement before it takes effect? | No reviewer, or a reviewer who rubber-stamps |
| Legal or similarly significant effect | Does it affect rights, contracts, money, access or opportunities? | Loan refusal, account closure, shortlisting rejection |

## When is human review "meaningful"?

The [WP29 guidelines on automated decision-making (WP251rev.01)](https://www.edpb.europa.eu/endorsed-wp29-guidelines_en), endorsed by the European Data Protection Board, say a human must have the authority and competence to change the decision, and must consider all relevant data. A token gesture does not count.

For agents, signs that review is not meaningful:

- the reviewer approves almost every recommendation with no recorded reasoning
- the reviewer sees only the agent's conclusion, not the inputs or sources
- volumes make real review impossible in the time allowed
- the interface defaults to "approve"

In December 2023, the Court of Justice of the EU ruled in the SCHUFA case (C-634/21) that generating a credit score can itself be an automated decision where a third party draws strongly on it to decide. The lesson for agents: a "recommendation" that downstream users almost always follow can be treated as the decision.

## What are the three exceptions?

Article 22(2) allows solely automated, significant decisions only if the decision is:

1. **Necessary for a contract** between the person and the controller.
2. **Authorised by EU or Member State law** that also sets suitable safeguards.
3. **Based on the person's explicit consent.**

For exceptions 1 and 3, Article 22(3) requires at least the right to obtain human intervention, to express their point of view and to contest the decision. Article 22(4) adds tighter limits where special category data, such as health data, is involved.

Articles 13(2)(f), 14(2)(g) and 15(1)(h) also require you to tell people about the existence of automated decision-making and give meaningful information about the logic involved.

## What evidence should you keep for agent decisions?

Use this checklist per agent that touches decisions about people:

1. **Classification.** A recorded decision on whether the agent's output is a decision, solely automated and significant, with the reasoning.
2. **Exception.** If Article 22 applies, which exception you rely on and why.
3. **Human review design.** Who reviews, what they see, their authority to override, and expected time per case.
4. **Override rate.** How often reviewers change the agent's recommendation. A rate near zero needs explaining.
5. **Decision log.** Per decision: agent version, inputs, sources retrieved, output, reviewer and final outcome.
6. **Contest route.** How a person requests human intervention, and records of those requests.
7. **Transparency notice.** The text shown to people, with the date and agent version it covers.
8. **Quality checks.** Evals on fairness and accuracy for the decision type, re-run when the agent changes.

[Evidence for human review of AI agent outputs](/blog/evidence-for-human-review-of-ai-agent-outputs) covers items 3 and 4 in more depth.

## How is Article 22 different from US automated decision rules?

US rules are converging on similar ideas but differ in detail. California's rules on automated decision-making technology focus on notice, opt-out and access rights for significant decisions. See [California ADMT rules and AI agents](/blog/california-admt-rules-and-ai-agents). Article 22 is framed as a prohibition with exceptions, which makes the "solely automated" question decisive.

## How Opencontroller does this

Opencontroller's governed registry records each agent, version and owner, and its connectors and runtime adapters capture traffic so decisions can be traced to a specific agent version. Live groundedness and task-adherence guardrails and LLM judges score outputs, and policy as code can route certain decision types to a human outside the agent. Opencontroller does not provide a human-approval hold in its gateway; human review runs in your workflow tools. These produce evidence for Article 22 safeguards. See [what an AI agent control plane is](/blog/what-is-an-ai-agent-control-plane) or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Is an AI agent recommendation an automated decision under GDPR?**
Not if a competent human with authority meaningfully reviews it. If the recommendation is almost always followed without real review, it may be treated as the decision.

**What safeguards does Article 22 require?**
Where the contract or consent exception is used, at least the right to human intervention, to express a view and to contest the decision.

**Does Article 22 ban AI agents from making decisions?**
No. It restricts solely automated decisions with legal or similarly significant effects to three exceptions, with safeguards.

## Key facts

GDPR Article 22: right not to be subject to decisions based solely on automated processing with legal or similarly significant effects. Exceptions: contract necessity, EU/Member State law, explicit consent; safeguards include human intervention, expressing a view and contesting. CJEU SCHUFA (C-634/21, Dec 2023): a score relied on heavily by a third party can be an automated decision. Meaningful human review needs authority and competence to change the outcome. Opencontroller provides registry, traceability, guardrails, LLM judges and policy as code as evidence; it has no gateway human-approval hold. Pricing starts from $250K per deployment per year.

## Related

- [How do you run a GDPR DPIA for an AI agent?](https://opencontroller.lyzr.com/blog/dpia-for-ai-agents.md)
- [What do California's ADMT rules mean for AI agents that make significant decisions?](https://opencontroller.lyzr.com/blog/california-admt-rules-and-ai-agents.md)
- [What evidence should you keep when a human approves an AI agent's output?](https://opencontroller.lyzr.com/blog/evidence-for-human-review-of-ai-agent-outputs.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
