AI licensing is an agreement that lets an AI company use your content or data, for training, testing or answering questions, in exchange for a fee, while you keep ownership. Licensed material includes articles, books, images, video, code and company records. Deals range from one-time training fees to usage-based and revenue-share terms.
- An AI license grants defined rights to use a copy; it is not a sale of the data.
- Content owners, platforms and ordinary companies with distinctive records all license to AI developers.
- Deal shapes are moving from one-time training fees toward usage-based terms.
- The permitted use, term and exclusivity matter as much as the fee.
What AI licensing is
In this guide, AI licensing means licensing data or content to AI developers, not licensing an AI product to customers. The owner grants a defined right to use a copy for stated purposes, such as training a model, testing it or retrieving passages to ground its answers, for a period and a fee. Ownership stays with the licensor.
For the clause-by-clause view of a company data license, read how AI data licensing works. For the difference between licensing and selling, see licensing vs selling data.
What gets licensed: content, records, code and video
Personal information is the common limit across every category. Licensors usually remove or de-identify it before a copy leaves the company.
| Category | Typical licensors | What AI developers use it for | Read more |
|---|---|---|---|
| Published content | News publishers, book and academic publishers, image libraries | Training, and grounding answers in current reporting | Deals tracker |
| Company records | Ordinary companies, and closing or bankrupt companies selling archives | Support histories, documents and internal communication to train and test assistants and agents | License company data |
| Source code | Software companies and closed startups | Training and evaluating coding models | Sell source code for AI training |
| Video and spatial data | Footage owners, survey firms, camera networks | Training robotics and world models | Selling company video footage |
| User-generated content | Online platforms and forums | Training on conversation and question-and-answer data | Deals tracker |
Deal shapes
Early AI licenses were mostly one-time fees for training. That is changing. Davis+Gilbert, a law firm, notes that publishers and AI platforms are starting to move past one-time training deals toward usage-based agreements, often called "grounding" or "RAG" (Davis+Gilbert).
| Shape | How payment works | When it fits |
|---|---|---|
| One-time training license | A fixed fee for a defined copy, used for training within the agreed scope | Archives that will not grow, such as a closed company's records |
| Multi-year license | An annual fee for access to an archive and new material over a term | Publishers and platforms that keep producing content |
| Usage-based or grounding license | Fees tied to how often the content is used in answers | Current content that AI products cite or retrieve |
| Revenue share | A portion of the buyer's subscription revenue or performance-based pay; Quartz reports these models are on the rise (Quartz) | Pooled catalogues where usage can be measured |
| Data access agreement | An initial fee for access to a defined collection | First deals where both sides are still testing value |
Who signs an AI license
The licensor is the party that holds the rights: a publisher, a platform or a company that owns its records. Inside a company, an owner or officer signs after legal and privacy review. Where customer or employee data is involved, the company must also be able to authorize that use under its contracts and privacy duties.
The licensee is an AI developer or a data company that prepares data for developers. Between them there is often a data partner, a marketplace or, in a shutdown, a wind-down firm. In a wind-down, MIT Technology Review reports that proceeds typically flow to senior creditors and investors, not employees, and that sellers often have little visibility into who is buying (MIT Technology Review).
Origin Data Partners is an independent data partner that helps companies get their data bought. A buyer may pay Origin when a deal closes.
What recent deals show
The AI data licensing deals tracker names the parties and sources for every public deal Origin has verified. Described without names, the pattern looks like this:
- A social platform disclosed data licensing contracts with an aggregate value of $203 million and terms of two to three years (Quartz).
- An academic publisher signed a $10 million initial data-access agreement (Quartz).
- A closed startup netted roughly $25,000 for its code base, with about $15,000 likely for its workspace data (MIT Technology Review).
- Suppliers selling shutdown archives employ data scientists to remove personal information and exclude HR records before sale (MIT Technology Review).
Reading the deals
Prices spread widely. Quartz reports that rates vary by orders of magnitude depending on the content, the publisher and the buyer (Quartz). Headline publisher deals cover vast archives and continuing content; a single company's records usually price far lower.
Terms keep evolving. Davis+Gilbert notes that deals depend on the negotiating strength of each party, and that AI platforms are learning which additional permitted uses to write into agreements (Davis+Gilbert). Read every permitted use before you sign.
To estimate what your own records could be worth, use the data valuation methods, then check fit for a rough range. Published ranges are not offers.
Common questions
What is AI licensing?
An agreement that lets an AI developer use your content or data for defined purposes, such as training, testing or grounding answers, for a fee and a period, while you keep ownership.
What kinds of data are licensed to AI companies?
Published content such as news, books and images; user content from platforms; source code; video and spatial data; and company records such as support histories, documents and internal communication.
How are AI licensing deals paid?
As one-time fees, annual fees over a term, usage-based fees or revenue shares. Davis+Gilbert notes a move from one-time training deals toward usage-based grounding agreements (Davis+Gilbert).
Who signs an AI data license?
The licensor that holds the rights, such as a publisher, platform or company, and the licensee, an AI developer or data company. Inside a company, an owner or officer signs after legal and privacy review.
How much do AI licensing deals pay?
It varies by orders of magnitude, according to Quartz (Quartz). Large publishers have reportedly licensed content for millions of dollars a year; a closed startup's code base sold for about $25,000. See the deals tracker for named, sourced deals.
Is an AI license the same as selling data?
No. A license grants defined rights to use a copy and you keep ownership; a sale transfers the data. Read licensing vs selling data for the details.
Sources and further reading
Origin Data Partners is independent and not affiliated with any buyer named on this page.