Class an AI use against the Act, article by article.

Software that only follows rules people wrote falls outside the Act (Article 3(1)).

Article 5(1): each is prohibited.

• Uses subliminal, manipulative or deceptive techniques that distort people's decisions in a way likely to cause significant harm
• Exploits people's vulnerabilities (age, disability, social or economic situation) to distort their behaviour in a way likely to cause significant harm
• Generates or alters realistic intimate or sexually explicit images, video or audio of an identifiable person without their explicit consent
• Generates or alters child sexual abuse material
• Scores people over time on their social behaviour or personal traits, leading to unjustified or unrelated detrimental treatment
• Predicts whether a person will commit a crime from profiling or personality traits alone
• Builds or expands facial recognition databases by untargeted scraping of images from the internet or CCTV
• Infers the emotions of people at work or in education (outside medical or safety reasons)
• Categorises people from biometric data to infer race, political opinions, trade union membership, religion, sex life or sexual orientation
• Real-time remote biometric identification in public spaces for law enforcement
• None of these
• Not sure

Article 6(1) and Annex I. Help features, efficiency and quality control are outside it unless a failure would endanger health or safety.

Annex III: the uses the Act lists as high-risk.

• Biometrics

  Remote biometric identification (verification that a person is who they claim to be is excluded)
• Biometric categorisation by sensitive or protected attributes
• Emotion recognition

• Critical infrastructure

  A safety component in managing critical digital infrastructure, road traffic, or the supply of water, gas, heating or electricity

• Education and vocational training

  Deciding access or admission to education or training, or assigning people to it
• Evaluating learning outcomes
• Assessing the level of education a person will receive or can access
• Monitoring students for prohibited behaviour during tests

• Employment

  Recruiting or selecting people: targeting job ads, filtering applications, evaluating candidates
• Decisions on promotion, termination or terms of work, allocating tasks by behaviour or traits, or monitoring and evaluating staff

• Essential services

  Evaluating eligibility for public assistance benefits and services, for or on behalf of a public authority
• Evaluating the creditworthiness of people or setting their credit score (financial fraud detection is excluded)
• Risk assessment and pricing for people in life or health insurance
• Evaluating emergency calls, dispatching emergency services, or emergency patient triage

• Law enforcement

  Use by or for law enforcement authorities (victim risk, polygraphs, evidence reliability, offending risk, profiling)

• Migration and border control

  Use by or for public authorities in migration, asylum or border control

• Justice and democratic processes

  Assisting a judicial authority, or alternative dispute resolution, in researching and applying the law to facts
• Influencing the outcome of an election or referendum, or how people vote
• None of these
• Not sure

Article 50: the transparency duties.

• People interact with it directly: a chatbot, a voice agent, an assistant replying to customers
• It generates synthetic audio, images, video or text
• It recognises emotions or categorises people from biometric data
• It makes or alters images, audio or video that resemble real people, places or events (a deep fake)
• It writes text published to inform the public on matters of public interest, without human review or editorial control
• None of these