How does Complipath check an AI system against the Article 5 prohibited practices?

What it does

• A question on every run covers the three practices that turn on a judgement. A tick sends the case to legal review; it never gives a prohibited verdict on its own.
• The fact-based practices are follow-ups, asked once you pick the matching area: criminal-risk profiling after law enforcement; face scraping and real-time identification after remote identification; emotion inference and sensitive categorisation after their own biometric options.
• A fact-based yes gives a prohibited verdict. The record gets no requirements checklist, because a prohibited practice cannot be made compliant.
• Exceptions are shown, not applied. The verdict text names the exception to criminal-risk profiling, the carve-out to sensitive categorisation and the objectives for real-time identification. The answer on emotion inference carries its own medical-or-safety exception.
• A question on generated sexual material, asked when the system produces content, covers the two practices Regulation (EU) 2026/1744 added. Before 2 December 2026 a matching answer is listed as a provision that starts later and sets no risk level.
• When a judgement-based practice is ticked, the page still shows what the other answers give with the Article 5 check set aside, and says the record stays open until a person answers it.

What it does not do

• It works from your answers and does not inspect the system. A practice you do not tick is not checked.
• It asks the fact-based practices only after you pick the matching area. A system placed in another area is not asked about them.
• It makes no legal judgement. Whether behaviour is materially distorted, harm is significant, or treatment is unjustified or disproportionate is left to legal review.
• A prohibited verdict is a flag, not a ruling. Each verdict text tells you to have counsel review it before you act.

Which plan

Every plan, including the free first system.

Where it sits in the law

• Article 5(1), first subparagraph, points (a), (b) and (c): subliminal, manipulative or deceptive techniques; exploiting vulnerabilities due to age, disability or a social or economic situation; and evaluating or classifying people over time by social behaviour or characteristics, where the social score leads to the unfavourable treatment the point describes.
• Each of the three carries conditions, such as materially distorting behaviour, significant harm, or unjustified or disproportionate treatment. These are the judgements the engine does not make.
• Article 5(1), first subparagraph, point (d): risk assessments to assess or predict the risk of a person committing a criminal offence, based solely on profiling or personality traits. It does not apply to systems that support a human assessment already based on objective and verifiable facts directly linked to a criminal activity.
• Article 5(1), first subparagraph, point (e): systems that create or expand facial recognition databases through untargeted scraping of facial images from the internet or CCTV footage.
• Article 5(1), first subparagraph, point (f): inferring emotions in the areas of workplace and education institutions, except where the system is intended for medical or safety reasons.
• Article 5(1), first subparagraph, point (g): biometric categorisation that deduces or infers race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation. It does not cover labelling or filtering of lawfully acquired biometric datasets, or categorising biometric data in law enforcement.
• Article 5(1), first subparagraph, point (h): 'real-time' remote biometric identification in publicly accessible spaces for law enforcement, unless and in so far as strictly necessary for one of three objectives. Article 5(2) to (6) set the further rules for such use.
• Article 5(1), first subparagraph, points (ba) and (bb) and Article 5(1a) and (1b) were inserted by Regulation (EU) 2026/1744. Article 5(1b) says what does not count as manipulation for point (ba). Read the provisions for their conditions and exceptions; this page does not restate them.
• Article 5(1a), point (a): placing on the market or putting into service is prohibited only where (i) generating or manipulating the material is the intended purpose, or (ii) the system's design, training, architecture, capabilities or user-facing functionalities make it a reasonably foreseeable and reproducible outcome without requiring significant technical modification, and the system lacks reasonable and adequate technical safety measures and other safeguards.
• Those safeguards are the ones that reliably prevent the generation or manipulation, taking into account reasonably foreseeable misuse, and correct observed or reported misuse. Point (b) covers use, where the deployer uses the system for that purpose.
• The assessment treats the intended purpose as a fact that gives a prohibited verdict, and the foreseeable outcome as a judgement that goes to legal review.
• Points (a), (b), (c), (d), (e), (f), (g) and (h) have applied since 2 February 2025 (Article 113, third paragraph, point (a)). Points (ba) and (bb), and Article 5(1a) and (1b), apply from 2 December 2026 (Article 113, third paragraph, point (a), as replaced by Regulation (EU) 2026/1744).