1What is it?
The EU AI Act is the world's first comprehensive legal framework regulating artificial intelligence systems across the European Union, adopted in 2024. It classifies AI applications into four risk tiers — unacceptable, high, limited, and minimal — and imposes obligations on developers and deployers accordingly. Concretely, it bans certain AI uses outright (like social scoring by governments), and places strict requirements on high-risk systems used in hiring, credit scoring, education, law enforcement, and healthcare. It applies to any company worldwide that deploys AI products targeting EU citizens, giving it global reach. Critics note the final text was significantly softened after intense lobbying by tech industry and some member states.
2What is it officially for?
The official objective is to ensure AI systems placed on the EU market are safe, transparent, and respect fundamental rights while fostering innovation and EU competitiveness. The Commission frames it as a 'human-centric' approach to AI governance that builds public trust. It aims to create legal certainty for businesses and prevent a fragmented patchwork of national AI rules. Notably, the stated goal of protecting citizens coexists with explicit carve-outs for national security and significant exemptions for law enforcement — areas where civil liberties organisations warn the protections are weakest. The innovation-protection balance was fiercely contested, with many NGOs arguing the final text prioritises industry interests over citizen rights.
3What it can change for citizens
Citizens will gain new rights including the ability to file complaints about high-risk AI systems affecting them, and companies must provide transparency about automated decisions in sensitive domains like job recruitment or loan approvals. Biometric mass surveillance in public spaces is nominally restricted, though with broad law enforcement exceptions that critics say render the ban largely symbolic. Workers assessed by algorithmic management systems, students evaluated by AI proctoring, and people scored by predictive policing tools will theoretically have more recourse — but enforcement depends entirely on national market surveillance authorities, whose capacity and independence vary widely. The Act also does not cover AI systems already deployed before its entry into force for several years, meaning millions of citizens remain unprotected during a long transition period. Vulnerable populations most exposed to algorithmic discrimination — migrants, benefit claimants, job seekers — may paradoxically see the weakest protections due to public-sector exemptions.