Back to glossary

Inference

Çıkarım

D1

In this glossary’s operational sense, inference is the use of a deployed model to derive outputs from new inputs.

Review status: 2027-02-23

Technical explanation

In operation, it connects an input to outputs such as predictions, content, recommendations, or decisions.

Conceptual boundaries

This runtime sense differs from the memorandum’s broader use of inference, which can also refer to deriving a model during the build phase.

Provider-neutral example

When a trained classifier receives a new message and suggests billing, it is performing inference.

Limitations

This entry uses the runtime sense of inference: a deployed model derives outputs from new inputs.

Related concepts

Atomic claims and evidence

  1. 1.1OECD, Explanatory Memorandum on the Updated OECD Definition of an AI System
    Source
    OECD, Explanatory Memorandum on the Updated OECD Definition of an AI System
    Source role
    Authoritative source
    Exact locator
    AI model, p. 7; Inferring how to generate outputs, p. 8
    Supported claim
    In this glossary’s operational sense, inference is the use of a deployed model to derive outputs from new inputs.
    Last verification
    Review due
    Scope limitation
    The OECD memorandum also uses inference in the build phase when a model is derived from inputs or data.
  2. 2.1Regulation (EU) 2024/1689 (Artificial Intelligence Act)
    Source
    Regulation (EU) 2024/1689 (Artificial Intelligence Act)
    Source role
    Authoritative source
    Exact locator
    Article 3(33)
    Supported claim
    An AI system’s input data are data provided to or directly acquired by the system on the basis of which it produces an output.
    Last verification
    Review due
    Scope limitation
    The claim concerns the relation between input and output, not the correctness of the output.