---
title: "Foundation model"
description: "A model trained on broad data at scale and adaptable to a wide range of downstream tasks."
canonical: https://aigovernanceengineer.com/glossary/foundation-model
author: "Jorge García Aibar"
license: "CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)"
doi: https://doi.org/10.5281/zenodo.22956197
version: "0.5.0"
updated: 2026-09-24
---

# Foundation model

A model trained on broad data at scale and adaptable to a wide range of downstream tasks [1]. Its defects are inherited by every system built on it, so organisations that call or adapt one collect the provider's evidence and manage the pinned version as a change.

- Developed in: [ch. 11, Foundation models and GPAI](https://aigovernanceengineer.com/bok/ai-defined#foundation-models-and-gpai)
- Chapters: [ch. 11, AI Defined](https://aigovernanceengineer.com/bok/ai-defined)
- Contrast with: [GPAI](https://aigovernanceengineer.com/glossary/gpai) · [Frontier model](https://aigovernanceengineer.com/glossary/frontier-model)
- In the glossary chapter: https://aigovernanceengineer.com/bok/glossary#t-foundation-model

## Sources

[1] On the Opportunities and Risks of Foundation Models ("trained on broad data at scale"; defects inherited downstream; arXiv 2108.07258). Bommasani et al. (Stanford CRFM). 2021-08-16. https://arxiv.org/abs/2108.07258 (verified: primary)