model license
In plain wordsThe legal terms under which an AI model may be downloaded and used — for example, whether businesses may use it, whether changes may be shared, and which uses are banned.
A license governing the use, modification and redistribution of a trained model's weightsweights, n.: The learned numbers inside a neural network that set how strongly each part influences the next. The finished set of weights is, in effect, the trained model itself. and associated code, and often of its outputs and derived models. Three families dominate. Standard permissive open-source licenses applied to weights, such as Apache 2.0 and MIT, impose few conditions. Responsible AI Licenses (RAIL and OpenRAIL) allow free use and redistribution but attach behavioral restrictions, such as bans on unlawful surveillance or on giving medical advice, that must pass to derivatives. Custom 'community' licenses, such as those for Meta's LlamaLlama, prop. n.: A family of AI language models from Meta, first released in 2023, whose trained settings can be downloaded and run by anyone, making them the foundation for thousands of other AI projects. models, add acceptable-use policies, attribution and naming requirements, and commercial thresholds; Llama 2's license, for example, required companies with more than 700 million monthly active users to seek separate permission. Some licenses also forbid using outputs to train competing models. Whether model weights are protected by copyright, and hence how enforceable such licenses are, remains unsettled in most jurisdictions, and use-restricted licenses do not meet the Open Source Definition.
Check the model license before shipping: the base model forbids commercial use, even if its fine-tunes don't say so.
Origin
Early research models were released under ordinary software licenses or research-only terms. The Responsible AI Licenses initiative, begun in the late 2010s, proposed behavioral-use clauses, adopted by BigScience for the BLOOM model and by Stability AIStability AI, prop. n.: A London company that paid for and released Stable Diffusion, a free image-generating AI anyone could download, and later models for video, audio and 3D. for Stable DiffusionStable Diffusion, prop. n.: A widely used image-generating AI whose model files were released for anyone to download, so people could run it on their own computers and adapt it. (CreativeML OpenRAIL-M), both in 2022; Meta's Llama 2 Community License (July 2023) became the template for custom commercial licenses.
Usage note
A model's license may differ from those of its code, training data and outputs, and fine-tuned derivatives usually inherit the base model's terms. License fields on hosting platforms are supplied by uploaders and are sometimes wrong.
Related
- open-weight modelopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private.
- open-source AIopen-source AI, n.: AI released, ideally with its code and details of its training data and not just the finished model, so that anyone may use, inspect, alter, share and even rebuild it. How open is open enough is debated.
- data licensingdata licensing, n.: The legal permissions that say who may use a dataset or a body of content, for what purposes and on what terms — including whether it may be used to train AI.
- Hugging FaceHugging Face, prop. n.: A company whose website is the main public library for sharing AI models and datasets — often called the GitHub of machine learning — and which makes widely used free AI software.
- LlamaLlama, prop. n.: A family of AI language models from Meta, first released in 2023, whose trained settings can be downloaded and run by anyone, making them the foundation for thousands of other AI projects.
- AI copyright litigationAI copyright litigation, n.: The many lawsuits in which authors, artists, news organizations, record labels and others accuse AI companies of copying their work without permission to build or run AI models.
Mentioned in
Hugging FaceHugging Face, prop. n.: A company whose website is the main public library for sharing AI models and datasets — often called the GitHub of machine learning — and which makes widely used free AI software., open-source AIopen-source AI, n.: AI released, ideally with its code and details of its training data and not just the finished model, so that anyone may use, inspect, alter, share and even rebuild it. How open is open enough is debated., open-weight modelopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private., proprietary modelproprietary model, n.: An AI model kept private by the company that made it; outsiders can use it only through the company's apps or paid online services and can never download it., data licensingdata licensing, n.: The legal permissions that say who may use a dataset or a body of content, for what purposes and on what terms — including whether it may be used to train AI., usage policyusage policy, n.: A company's published rules on what people may and may not use its AI for — for example, no help building weapons, no scams and no sexual content involving children., Open Source InitiativeOpen Source Initiative, prop. n.: A nonprofit group that decides what counts as 'open source' software and, in 2024, published a definition of what makes an AI system genuinely open source.