Article 5(1)(a) prohibits the placing on the market, the putting into service or the use of an AI system that deploys subliminal techniques beyond a person's consciousness, or purposefully manipulative or deceptive techniques, with the objective or the effect of materially distorting a person's or a group of persons' behaviour by appreciably impairing their ability to make an informed decision, thereby causing that person to take a decision that they would not have otherwise taken in a manner that causes or is reasonably likely to cause that person, another person or a group of persons significant harm. Every element must be present. Ordinary persuasion, advertising and design that leave informed decision-making intact are not described by this wording, but the boundary is fact sensitive.
What techniques does point (a) describe?
Point (a) describes two kinds of technique. Subliminal techniques operate beyond a person's consciousness, so the person is not aware of the influence being applied. Purposefully manipulative or deceptive techniques are designed to mislead or to steer behaviour by means that undermine informed choice. The word purposefully matters: the technique must be deployed with that character, not merely have an incidental effect that was neither designed nor reasonably foreseeable in the relevant sense. A record should describe the technique itself, how it operates, what the person sees or hears, and what the designer intended it to do, rather than relying on a general claim that a feature is engaging or persuasive.
What does materially distorting behaviour mean?
Material distortion means the behaviour is changed in a way that matters, by appreciably impairing the person's ability to make an informed decision. The impairment must be appreciable, not trivial, and it must affect the decision-making ability itself, not merely present information in a particular order. A system that presents options clearly, discloses material facts, and leaves the person free to choose differently does not meet that description on its face. A system that hides material information, presents false premises as true, exploits a lack of awareness of the technique, or creates pressure that bypasses reflection needs closer examination. The test is applied to the person or group affected, in their actual context, not to an abstract reasonable user.
For the broader regulatory framework, see prohibited practices under Article 5.
How is significant harm assessed?
The distortion must cause, or be reasonably likely to cause, significant harm to the person, another person, or a group of persons. Significant harm is not limited to financial loss. It can include harm to health, safety, autonomy and other interests, assessed in the context of the decision taken. The harm must be linked to the decision that the person would not otherwise have taken. A speculative or remote harm, or a harm unconnected to the distorted decision, does not meet the wording. Because harm is assessed prospectively as reasonably likely as well as retrospectively as caused, a system should be screened before deployment, not only after harm is observed, and the screening should record the foreseeable decisions and their foreseeable consequences.
Where is the line with persuasion and marketing?
Persuasion that operates openly, presents true information, and respects the person's ability to decide differently is not the practice point (a) describes. Marketing, recommendation and personalisation that disclose their nature and leave the person able to make an informed decision fall on the other side of the line, although other laws, including consumer protection and data protection law, may still apply. The line moves when techniques are hidden, deceptive or designed to bypass awareness, when material facts are withheld or falsified, or when the design targets a person's inability to recognise the influence. Product teams should document the disclosure, the information provided, and the choices available, because those facts are central to showing where a feature sits.
What evidence should a screening record contain?
A screening record for point (a) should describe the user journey, the content presented, any hidden or subliminal element, the intended behavioural effect, the decision affected, and the foreseeable harm. It should include screenshots or transcripts where relevant, the design rationale, and any testing of user understanding. If the record concludes that the point is not met, it should state which element failed, for example that no subliminal or purposefully manipulative technique was deployed, that decision-making ability was not appreciably impaired, or that significant harm was not caused or reasonably likely. A bare statement that a feature is not manipulative does not evidence the test.
Continue with exploitation of vulnerability.
What does this guide leave for qualified review?
This guide explains the elements of point (a) in outline. It does not decide whether any particular design, prompt, interface or campaign is prohibited. That decision needs the full technique, the actual user experience, the decisions affected, and the harm analysis, read against the consolidated text. Behavioural design is fact sensitive, and small changes in disclosure or pressure can change the analysis. Where a feature is designed to influence consequential decisions, or where user testing shows misunderstanding or pressure, confirmation should be sought from a competent authority or legal adviser before the feature is placed on the market, put into service or used.
Frequently asked questions
Is all personalisation prohibited by Article 5(1)(a)?
No. Personalisation must meet every element of point (a), including a subliminal or purposefully manipulative or deceptive technique, material distortion of behaviour through appreciable impairment of informed decision-making, and significant harm or its reasonable likelihood.
Does point (a) require proof that harm already happened?
No. The wording covers causing significant harm and being reasonably likely to cause it. Screening should therefore be prospective, based on foreseeable decisions and consequences, as well as retrospective where harm has occurred.
Can a disclaimer cure a manipulative technique?
Not necessarily. A disclaimer is relevant evidence about awareness and informed decision-making, but it does not automatically remove a subliminal or purposefully manipulative technique, especially where the technique operates beyond consciousness or the disclaimer is not actually seen or understood.
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 ↗AI Act Service Desk timeline ↗European Commission regulatory framework for AI ↗