Skip to content

SourceX Research · public sources

Common terms in public data licensing agreements

Quick answer

Public reporting on AI data deals and openly published template licenses both tend to cover the same handful of issues: scope of permitted use, duration, exclusivity, and whether the data can be redistributed or must stay with one licensee. This page reviews publicly available templates and press-reported deal terms; it summarizes public sources and is general information, not legal advice.

By Noah Loul · Updated

This report summarizes published public sources, listed at the end. It does not contain SourceX survey or deal data.

License terms

Common terms in public data

This page covers common terms in public data licensing agreements.Illustrative

Open template licenses built for data

The Linux Foundation's Community Data License Agreement (CDLA) exists specifically because open-source software licenses weren't designed for datasets; it offers a 'Sharing' variant (share-alike style, requiring downstream sharing of improvements) and a 'Permissive' variant that places fewer conditions on reuse, with a Version 2.0 of CDLA-Permissive released as a shorter, more streamlined agreement [1](https://cdla.dev/) [2](https://cdla.dev/faq-resources/faq/).

Creative Commons and Open Data Commons

Creative Commons licenses and the Open Data Commons family of licenses (such as the Open Database License) are long-standing public templates that specify attribution requirements, whether commercial use is allowed, and whether derivative datasets must be shared under the same terms. These are widely used for openly published datasets rather than negotiated commercial AI-training deals.

What press-reported AI training deals disclose

Public reporting on commercial AI data deals rarely discloses full contract text, but discloses some structural terms. Reddit's 2024 SEC Form S-1 disclosed that licensing contracts with large language model companies are structured as multi-year (two- to three-year) term deals providing 'continuous access' to Reddit's data API plus periodic data transfers, rather than a one-time data dump [3](https://arstechnica.com/ai/2024/02/reddit-has-already-booked-203m-in-revenue-licensing-data-for-ai-training/). Shutterstock's agreements with AI developers, including a six-year deal with OpenAI signed in 2023, are structured as multi-year terms covering images, video, music and associated metadata [4](https://investor.shutterstock.com/news-releases/news-release-details/shutterstock-expands-partnership-openai-signs-new-six-year).

Exclusivity and scope are usually limited

Public reporting indicates these deals are generally non-exclusive: Shutterstock has disclosed parallel licensing relationships with multiple AI companies including Meta, Alphabet, Amazon and Apple alongside OpenAI, meaning its content is not locked to a single Acquisition Partner [5](https://finance.yahoo.com/news/shutterstock-ai-licensing-business-generated-120000890.html). News Corp's 2024 deal with OpenAI was described by the companies as a multi-year content license giving OpenAI access to current and archived material, rather than a transfer of ownership [6](https://investors.newscorp.com/news-releases/news-release-details/news-corp-and-openai-sign-landmark-multi-year-global-partnership).

What this means for a business

A US company considering any kind of data license can use these public structures as reference points: term length (often multi-year, not perpetual), whether access is continuous/updating versus a one-time delivery, and whether the arrangement is exclusive or allows licensing to multiple parties. Before agreeing to any terms, a business would still need to check what it actually has the rights to share, know which records hold personal details that must be de-identified before use (under a SourceX agreement, SourceX does that, through approved processing routes), and decide what, if anything, should be excluded from scope.

Frequently asked questions

What is the CDLA?

The Community Data License Agreement is a set of Linux Foundation template licenses for sharing datasets, with 'Sharing' (share-alike) and 'Permissive' variants [1](https://cdla.dev/).

Are AI training data deals usually exclusive?

Public reporting suggests often not. Shutterstock, for example, has disclosed parallel deals with several AI companies rather than one exclusive Acquisition Partner [5](https://finance.yahoo.com/news/shutterstock-ai-licensing-business-generated-120000890.html).

How long do these deals typically run?

Reported terms vary: Reddit's SEC filing described two- to three-year licensing contracts, while Shutterstock's OpenAI deal is six years [3](https://arstechnica.com/ai/2024/02/reddit-has-already-booked-203m-in-revenue-licensing-data-for-ai-training/) [4](https://investor.shutterstock.com/news-releases/news-release-details/shutterstock-expands-partnership-openai-signs-new-six-year).

Is this legal advice on how to structure a license?

No. This is a summary of publicly reported deal structures and open license templates, not legal advice.

Sources

Public sources, checked 2026-10-09. SourceX is not affiliated with the organizations named.

  1. Linux Foundation's Community Data License Agreement offers Sharing and Permissive variants for datasets cdla.dev
  2. CDLA-Permissive 2.0 is a shorter, more streamlined data agreement cdla.dev
  3. Reddit's S-1 disclosed two- to three-year AI data licensing contracts with continuous API access plus data transfers arstechnica.com
  4. Shutterstock signed a six-year data licensing agreement with OpenAI in 2023 investor.shutterstock.com
  5. Shutterstock has parallel AI licensing customers including Meta, Alphabet, Amazon and Apple finance.yahoo.com
  6. News Corp and OpenAI's multi-year content deal gives OpenAI access to current and archived material investors.newscorp.com

See if your company qualifies

A short company assessment. No data uploads are needed.

See if your company qualifies