GDPR Article 22 · Practice

What counts as meaningful human involvement under Article 22?

Updated 6 July 2026 · 6 min read

Key takeaway

Not a rubber stamp. To take a decision out of "solely automated", the human must have genuine authority and real capacity to reach a different conclusion, and must actually consider the case, not merely sign off the machine's output. Token review does not escape Article 22, and the SCHUFA reasoning reinforces this.

Any human in the process does not automatically defeat the "solely automated" test.
Meaningful involvement requires authority, competence and information, and genuine consideration.
A review that never overturns anything is evidence of a rubber stamp.
The same oversight discipline satisfies the EU AI Act's human-oversight duty.
General information, not legal advice. Current as of July 2026.

The rubber-stamp trap

The most common Article 22 mistake is assuming that any human in the process defeats the "solely automated" test. Regulators are clear that it does not. A human who lacks the authority, information, or time to overturn the system's output, or who defers to it by default, adds no meaningful involvement, and the decision remains solely automated in substance.

What meaningful involvement requires

Three things. Authority: the reviewer must be empowered to reach a different outcome, not merely to approve. Competence and information: they must understand the decision and see the inputs, not just the score. Genuine consideration: they must actually weigh the individual case, with the capacity to depart from the recommendation. Designing review to satisfy all three is what moves a process out of Article 22's prohibition.

Designing it honestly

Build the reviewer's role so departures from the system are expected and recorded, give them the underlying facts rather than only the output, and measure how often human and machine diverge, because a review that never overturns anything is evidence of a rubber stamp. The same human-oversight discipline satisfies the EU AI Act's oversight duty, so design it once, to the higher bar, and let both regimes draw on it.

Key terms

Meaningful human involvement: Review that has real authority, information, and consideration behind it.
Genuine authority: The reviewer can reach a different outcome, not just approve the system.
Rubber stamp: Nominal review that defers to the automated output by default.
Override rate: How often human reviewers depart from the system — a signal of real involvement.
Human oversight: The EU AI Act's parallel duty to design oversight into high-risk systems.

References

Related guides