The EU AI Act is now law and it will affect products built and used outside Europe more than most people realise

The European Parliament's EU AI Act explainer https://www.europarl.europa.eu/topics/en/article/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence is the starting point for understanding what the first major AI law in the world actually does and why its effects extend well beyond Europe.

The risk-based approach is the core design principle. Systems classified as unacceptable risk, social scoring, certain biometric surveillance, cognitive behavioural manipulation, are banned outright. High-risk systems in categories like critical infrastructure, education, employment, healthcare and law enforcement face strict requirements around transparency, accuracy, human oversight and data governance. Generative AI systems have specific transparency obligations regardless of their risk classification.

The extraterritorial effect is the dimension most non-European coverage underweights. Any product used by people in the EU is subject to the Act regardless of where it is built or headquartered. That means US, Asian and Australian AI companies building products with European users are affected by the compliance requirements.

The implementation timeline being phased means some requirements are already active and others are still coming. The product design and compliance implications are already part of real decisions being made by AI companies now.

Will the EU AI Act improve user trust and safety or will it primarily slow European AI adoption relative to less regulated markets?