-
Notifications
You must be signed in to change notification settings - Fork 2
ModelLicensing
⚠️ Sample. An AI-generated demo of LLM Wiki Newsroom — the "open source AI" topic is just the example corpus.
Model licensing governs how an AI model may be used, modified, and redistributed, and it is where open-source AI diverges from traditional open-source software. Conventional OSS licenses split into permissive families (MIT, BSD, Apache 2.0) that impose few obligations and copyleft families (GPL 2.0/3.0, Affero GPL) that require sharing derived source under the same terms. AI models add components beyond software — training data, data information, weights, and parameters — so a software license alone cannot make a model open. DeepSeek R1 ships its weights under the permissive MIT license, while Meta's Llama uses a custom community license with usage restrictions that the OpenSourceInitiative judges incompatible with the OpenSourceAI freedoms.
- OpenSourceAI — licensing is one component the OSAID evaluates
- OpenWeights — a licensing posture that releases weights but not data
- OpenWashing — restrictive license terms enable open-washing
- OpenSourceInitiative — the body that judges license compatibility
- Meta — issuer of the contested Llama community license
- catalog-licensing-open-washing
- catalog-open-source-ai-definition
- catalog-open-weights
- catalog
- The Case Against OSI's Open Source AI Definition
- Celebrating an Important Step Forward for Open Source AI (Mozilla)
- Open Source AI Models: How Open Are They Really? (Part 1)
- The Open Source AI Definition 1.0