First proposed by the European Commission in April 2021, the European Parliament has agreed to the world’s first ever comprehensive legal framework on Artificial Intelligence (AI). This groundbreaking Act is a risk-based approach designed to protect citizens and guide AI developers and deployers. It must be formally adopted by both the Parliament and Council to become EU law and will come into full effect in 2026. The Act does not cover AI used for military/defence, research/innovation or individual, non-professional use. AI systems that manipulate human behaviour to circumvent a citizen’s free will or exploit vulnerable people are considered an ’unacceptable risk’ and are prohibited, including emotional recognition in the workplace, untargeted scraping of facial images from CCTV/internet to create facial recognition databases and Biometric identification. AI systems are ‘high risk if they could potentially risk a citizen’s life or health and are subject to a mandatory fundamental rights impact assessment (amongst other requirements) prior to approval. This includes AI systems for critical infrastructures, medical devices, the education system, recruitment, border control, the justice system, and biometric identification, categorisation and emotion recognition systems.
Law enforcement use of AI biometrics is limited to targeted searches for victims or specific criminals and prevention of specific terrorist attacks. General AI systems and models capable of quick expansion/development with a potential threat of systemic risks will be subject to additional binding obligations, requiring mandatory codes of practice. ‘Minimal risk’ AI systems (a category most current AI systems fall into) have no obligations, but companies can voluntarily apply codes of conduct for use. These are deemed to pose little to no risk to a citizen’s rights or safety (e.g. AI-enabled recommender systems or spam filters).
‘Specific transparency risk’ applies to AI systems such as chatbots which must be labelled as such to ensure that the user is aware that they are interacting with AI. Looking forward to the development and implementation, a new European AI office will be established to monitor and enforce, the first of its kind globally. National market surveillance authorities will supervise implementation at a national level, and the EU continues its work towards trustworthy AI at an international level with the OEDC, G7, G20 and UN. Read the European Commission’s press release:
https://ec.europa.eu/commission/presscorner/detail/en/IP_23_6473
