AI system and model
The OECD definition centres on a system inferring outputs from inputs, for objectives, with potential effects in physical or virtual environments. A model is a component used to produce inferences; the deployed system also includes the surrounding tools and operating context. [1]
Autonomy and adaptation
Autonomy concerns how far a system acts without human involvement after delegation. Adaptation concerns changes after initial development. A system can have substantial delegated access without continuously learning, so the two properties should not be treated as interchangeable. [1]
ADM perspective
Automation is the broader workflow question: which steps run without a person carrying them out? A rule that routes invoices can automate work without using a generative model. A language model that drafts an answer can support a decision without having authority to execute it. Ask separately what infers, what acts and who remains responsible. These examples are a practical aid, not a test of legal scope.
Read the original sources
The original publications provide the full detail, evidence and limitations.
- OECD · Intergovernmental explanatory memorandumExplanatory memorandum on the updated OECD definition of an AI system
An AI-assisted, source-led summary. “ADM perspective” is editorial interpretation, separate from the source’s findings. Our editorial approach.