What is open-source AI?

Open-source AI usually means AI models that anyone can download, run and modify. Most models described this way are strictly open-weight: the trained model is published, but the training data and most of the code are not. Under the Open Source Initiative’s definition, open-source AI must also share the training code and enough detail about the training data to build a similar model. Very few major models meet that standard.

Why it matters for a business

The difference shows up in contract review. An open-weight model under a permissive licence behaves like open-source software: use it, change it, ship it with attribution. One under the lab’s own licence may carry acceptable-use rules or size thresholds that need a separate agreement. Regulation can turn on it too. The EU AI Act eases some documentation duties for models released under a genuinely free and open licence, though not the copyright obligations, and not at all for the largest models.

Go deeper

Open-source vs. closed AI models: what is actually open

Related terms

• Open-weight model
• Private LLM
• Foundation model
• Training data

All terms, A to Z