EU Regulations on AI Use Prohibitions

Prohibited AI Practices in the EU

• Deploying AI systems with manipulative 'subliminal techniques' is prohibited.
• Exploiting specific vulnerable groups based on traits, social status, age, or disability is prohibited.
• Expanding facial recognition databases through untargeted web image scraping is prohibited.
• Infering emotions of a natural person in workplace and education institutions is prohibited.
• Using AI for social scoring is prohibited.
• Employing remote biometric identification in public spaces is prohibited.
• Fines for non-compliance can reach up to 7% of global revenues or 35 million euros, whichever is higher.

High-Risk AI Systems

• The AI Act defines 'high-risk' AI systems that threaten health, safety, or fundamental rights.
• Examples of high-risk AI systems include:
  • Use in health and safety products (e.g., toys, aviation, cars, medical devices, lifts).
  • Management and operation of critical infrastructure.
  • Access to essential private and public services and benefits (e.g., credit, healthcare, insurance).
  • Employment, worker management, and access to self-employment.
  • Biometric and biometrics-based systems (incl. non-prohibited emotion recognition systems).
  • Education and vocational training.
  • Influencing the outcome of an election or referendum.
  • Used in social media recommender systems for content recommendations.
  • Administration of justice and democratic processes.
  • Migration, asylum, and border control management.

Obligations for Deployers of High-Risk AI

• Fundamental Rights Impact Assessment is required.
• Diligent Use & Monitoring is required.
• Data Quality & Governance is required.
• Define purpose, scope, and impact.

AI Act Objective and Principles

• Objective: Ensure a well-functioning single market with reliable AI.
• Principles:
  • Enable a well-functioning single market with reliable AI.
  • Ensure that AI systems placed on the EU market are safe and respect existing EU law.
  • Enhance governance and effective enforcement of EU law on fundamental rights.
  • Ensure legal certainty to facilitate investment and innovation in AI.
  • Facilitate the development of a single market for lawful, safe, and trustworthy AI applications.
  • AI to improve individual and collective wellbeing.
  • Respect for human autonomy.
  • Fairness.
  • Humans must be able to keep full self-determination.
  • Individuals must be free from unfair bias or discrimination.
  • Prevention of harm.
  • Explicability.
  • Avoid harm and protect human dignity and well-being.
  • Transparent processes and purpose, explainable decisions.

AI Act Benefits and Risks

• Benefits:
  • Employee retention.
  • Competitive advantage.
  • Improved customer journey.
  • Improved efficiency and lower costs.
  • Trust, security, transparency.
  • Minimise unintended risks.
  • Improved decision-making.
• Risks:
  • Underuse and overuse of AI.
  • Liability.
  • Threats to fundamental rights and AI impact on jobs.
  • Competition.
  • Safety and security risks.
  • Transparency challenges democracy.

AI Act Key Elements

• AI system: A machine-based system that is designed to operate with varying levels of autonomy.
• Provider: Any individual or entity that creates or commissions an AI system for market deployment.
• Importer: A person or entity within the Union introducing an AI system to the market under a non-Union entity's name or trademark.
• Distributor: Any entity in the supply chain, excluding the provider and importer, making an AI system available in the Union market without altering its properties.
• Deployer: Any person, organization, or authority utilizing an AI system under their control, except for personal non-professional use.

Risk-Based Approach

• The EU AI framework adopts a risk-based approach, which lays down different requirements and obligations for the development, placing on the market, and use of AI systems in the EU.
• Examples of high-risk AI applications:
  • Social scoring.
  • Remote biometric identification.
  • Worker monitoring.
• Examples of low-risk AI applications:
  • Chatbots.
  • Content creation.
  • Back-office automation.
  • Video games.
  • Spam filters.