Explanatory memorandum on the updated OECD definition of an AI system (oecd.org)

The OECD published an explanatory memorandum updating its working definition of an “AI system,” clarifying how governments, regulators and organizations should identify what counts as AI for policy, procurement and oversight. The memorandum lays out a technology‑neutral, function‑focused framing that spans statistical and machine‑learning models, optimization and rule‑based systems, and the full model lifecycle (training, tuning, deployment and monitoring). It also emphasizes inputs and outputs, human oversight, and the ways systems produce inferences or automated decisions — giving practitioners a clearer baseline for when governance rules apply. This update matters because a common, operational definition underpins consistent risk‑based governance, cross‑border regulatory alignment and technical standards. By specifying scope and lifecycle touchpoints, the memorandum strengthens requirements around transparency (e.g., documentation and model cards), testing and validation, and accountability measures that developers and procurers will need to adopt. It also enables tighter interoperability with other instruments — from national AI laws to standards bodies and statistical reporting — shaping compliance, auditing, labelling, and data‑sharing practices across the AI/ML ecosystem.