EU AI Act

The EU AI Act is the European Union's regulation for artificial intelligence systems, built around a risk-based classification: AI applications are grouped into risk tiers, from minimal risk through high-risk to a small set of practices banned outright, such as certain forms of biometric categorization or social scoring, with obligations scaling to the tier a given system falls into. High-risk systems, which include many used in employment, credit, and essential public services, face requirements around documentation, human oversight, and risk management before and after deployment.

The Act differs from General Data Protection Regulation in what it regulates: GDPR governs personal data generally, while the AI Act targets AI systems by risk profile regardless of whether personal data is involved, though the two overlap heavily in practice. It formalizes many principles previously treated as voluntary Responsible AI practice into binding requirements for in-scope systems, closely tied to model risk management processes.

For analytics and data science teams, the Act matters because it can require formal risk assessments, technical documentation, and human oversight for models that would previously have been deployed with informal review, and because obligations apply based on how a system is used, not just where the deploying company is based. It's closely related to algorithmic bias testing and Explainable AI requirements. This is general background, not legal advice; applicability depends on how a given system is classified.