01 · FEIT
Social scoring verboden (Art. 5(1)(c))
De AI Act verbiedt het op de markt brengen, in gebruik nemen of gebruiken van AI-systemen die natuurlijke personen of groepen over een periode evalueren of classificeren op sociaal gedrag of (afgeleide) persoons-/persoonlijkheidskenmerken, wanneer de sociale score leidt tot nadelige behandeling in een andere sociale context en/of tot ongerechtvaardigde of onevenredige behandeling.
«the placing on the market, the putting into service or the use of AI systems for the evaluation or classification of natural persons or groups of persons over a certain period of time based on their social behaviour or known, inferred or predicted personal or personality characteristics, with the social score leading to either or both of the following: (i) detrimental or unfavourable treatment of certain natural persons or groups of persons in social contexts that are unrelated to the contexts in which the data was originally generated or collected; (ii) detrimental or unfavourable treatment of certain natural persons or groups of persons that is unjustified or disproportionate to their social behaviour or its gravity;»
02 · FEIT
High-risk: publieke bijstand (Annex III)
Naast product-safety high-risk (Art. 6(1)) zijn AI-systemen in Annex III high-risk (Art. 6(2)). Annex III punt 5(a) dekt systemen die overheden (of namens hen) gebruiken om te beoordelen of personen in aanmerking komen voor essentiële publieke bijstand/diensten (incl. zorg) en om die toe te kennen, te verlagen, in te trekken of terug te vorderen — AI als poortwachter van publieke rechten.
«AI systems intended to be used by public authorities or on behalf of public authorities to evaluate the eligibility of natural persons for essential public assistance benefits and services, including healthcare services, as well as to grant, reduce, revoke, or reclaim such benefits and services;»
03 · FEIT
GPAI systemic risk: >10²⁵ FLOPs
Een general-purpose AI-model wordt vermoed high-impact capabilities te hebben (en zo als GPAI met systemic risk te classificeren) wanneer de cumulatieve trainingsberekening gemeten in floating point operations groter is dan 10²⁵ (Art. 51(2)). Drempel als besturingsknop voor de grootste modellen.
«A general-purpose AI model shall be presumed to have high impact capabilities pursuant to paragraph 1, point (a), when the cumulative amount of computation used for its training measured in floating point operations is greater than 10²⁵.»
04 · FEIT
AI Office: exclusieve GPAI-handhaving
De Commissie heeft exclusieve bevoegdheden om Chapter V (GPAI) te supervisieren en handhaven; die taken vertrouwt zij toe aan het AI Office (Art. 88(1)). Institutionele poort van het besturingskader — niet alleen soft policy-pagina.
«The Commission shall have exclusive powers to supervise and enforce Chapter V, taking into account the procedural guarantees under Article 94. The Commission shall entrust the implementation of these tasks to the AI Office, without prejudice to the powers of organisation of the Commission and the division of competences between Member States and the Union based on the Treaties.»
05 · FEIT
Biometrische categorisatie verboden (Art. 5(1)(g))
Los van realtime remote biometrie (LFR-knoop) verbiedt Art. 5(1)(g) biometrische categorisatiesystemen die personen individueel categoriseren om ras, politieke opvattingen, vakbondslidmaatschap, geloof, seksleven of seksuele geaardheid af te leiden. Andere snede dan LFR en dan het al soft vastgelegde emotie-verbod 5(1)(f).
«the placing on the market, the putting into service for this specific purpose, or the use of biometric categorisation systems that categorise individually natural persons based on their biometric data to deduce or infer their race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation; this prohibition does not cover any labelling or filtering of lawfully acquired biometric datasets, such as images, based on biometric data or categorizing of biometric data in the area of law enforcement;»