Skip to content

AI Model Licences Explained

For people

Apache-2.0, MIT, OpenRAIL, community licences and more: what common model licences allow and what to check before using a model.

Editorial team 2 min read

Model weights are released under many different licences, and "open" rarely means "no conditions". Check the licence before building on a model.

Permissive Open-Source Licences

  • Apache-2.0 and MIT allow use, modification and commercial distribution with few conditions, mainly keeping copyright and licence notices. Apache-2.0 also includes an explicit patent grant.

Responsible AI Licences

  • OpenRAIL and similar licences permit broad use but add use restrictions — prohibiting certain harmful applications — which must be passed on to anyone you distribute the model to.

Custom Community Licences

Some widely used open-weight models ship with bespoke licences that may:

  • require attribution in a specific form;
  • restrict use above a certain number of users without a separate agreement;
  • prohibit using outputs to train competing models;
  • include an acceptable-use policy.

Non-Commercial Licences

Some models are released for research or non-commercial use only.

What to Check

  • Is commercial use allowed?
  • Are there user, revenue or deployment limits?
  • Can you fine-tune and redistribute modified versions?
  • Are there restrictions on using outputs?
  • What attribution is required?
  • What licences apply to the training data and any datasets you combine with it?

Keep Records

Record the model, version and licence used in each product, and review terms when upgrading. For significant commercial use, get legal advice — this guide isn't legal advice.

More in Generative AI

All Generative AI guides →
Generative AI Guide · 2 min

Prompt Engineering Fundamentals

The building blocks of a good prompt — context, task, constraints and format — with before-and-after examples.

Generative AI 2 min read 25 Jul 2026 For people

Generative AI Guide · 2 min

Few-Shot Prompting With Examples

Showing a model a few examples of the input and output you want is often clearer than describing it. How to choose good examples.

Generative AI 2 min read 24 Jul 2026 For people

Generative AI Guide · 2 min

Getting Structured Output From LLMs

How to get JSON and other machine-readable output reliably from a language model, and how to validate it.

Generative AI 2 min read 23 Jul 2026 For people

Generative AI Guide · 2 min

Why Language Models Hallucinate

What hallucination is, why it happens, and practical ways to reduce and catch it.

Generative AI 2 min read 22 Jul 2026 For people

Share X LinkedIn Reddit Bluesky Email

Discussion 0

Use @handle to bring somebody in, and #tag to file it.

Nothing here yet. Ask a question, or say how you used it.

Sign in to join the discussion. Sign in