GovernanceCore
Foundations

Foundation model

Working definition

Reviewed 30 July 2026

A foundation model is a model trained on broad data—generally using self-supervision at scale—that can be adapted to a wide range of downstream tasks. The term emphasizes that the model is a shared but incomplete base on which many applications are built.

Context

Why it matters

Foundation-model dependencies concentrate risk, complicate assurance, and create information asymmetry between model developers and downstream deployers.

Operating note

What this looks like in practice

  1. 01Record model version, provider, terms, evaluations, limitations, and downstream dependencies.
  2. 02Evaluate the complete system in its use context rather than relying only on upstream benchmarks.
  3. 03Plan for model updates, withdrawal, access changes, and inherited incidents.

Sources & further research

Primary authority anchors the definition. Research links add conceptual or operational depth. External sources may update independently; always verify legal duties against the current official text.

  1. Research paper

    On the Opportunities and Risks of Foundation Models

    Introduces the term and analyzes capabilities, homogenization, downstream impacts, and governance challenges.

    Stanford Center for Research on Foundation Models
    2021    Open source ↗
  2. Official source

    Artificial Intelligence Risk Management Framework: Generative AI Profile

    Profiles distinctive generative-AI risks and suggested risk-management actions.

    NIST
    2024    Open source ↗