Industry
Media and publisher data licensing for AI
Outrich Data licenses text, audio and video archives from publishers and media companies with 50 or more employees and supplies them to AI labs for training, fine-tuning and evaluation. Only works you own or control are included. You keep the copyright, approve each package in writing, and are not named to buyers without consent.
Media and Publishers data and AI labs
A publisher's archive is edited, fact-checked and dated. That makes it more useful to an AI lab than the same volume of unedited web text. Trade publications, regional news, specialist magazines, book lists and broadcast libraries all carry subject depth that general sources lack.
The main question in media is not privacy. It is ownership. An archive usually mixes staff work the company owns outright with freelance pieces, wire copy, stock images and syndicated columns that came in under narrower terms. A license can only cover what you have the right to grant.
Outrich Data works with media companies that have at least 50 employees and two years of operating history. The archive is sorted by rights status first. Works with unclear rights are held back until the contributor agreement is found or the contributor agrees.
Records in demand
| Record | What AI labs use it for |
|---|---|
| Published article archives with dates and sections | Training on edited prose with reliable metadata about topic, time and desk. |
| Edit histories and revision drafts | Teaching models what an editor changes between a first draft and the published version. |
| Fact-check notes and corrections logs | Evaluating whether a model can catch the kinds of errors a research desk catches. |
| Broadcast and podcast audio with transcripts | Training speech and summarization models on professionally produced spoken content. |
| Video libraries with rundowns and captions | Training video understanding models on footage paired with accurate human descriptions. |
| Photo archives with captions and credits | Training image and caption models, limited to images where the publisher holds the rights. |
| Style guides and editorial standards | Fine-tuning models to write to a consistent house standard. |
Rules that apply
What the title search looks at in media and publishers.
- Copyright ownership and work made for hire
- Under the US Copyright Act, an employer generally owns works employees create within the scope of their jobs. Work by independent contractors is owned by the contractor unless a written agreement says otherwise. Review ownership by contributor type with counsel.
- Freelancer and contributor agreements
- Many older agreements grant print or first publication rights only. The Supreme Court's decision in New York Times Co. v. Tasini confirmed that such grants do not automatically extend to new electronic uses. Each agreement era should be read.
- Wire, syndication and stock licenses
- Content from wire services, syndicates and photo agencies is licensed to you for publication, not for relicensing. It needs to be identified and removed from any package.
- Talent and guild agreements, and right of publicity
- On-air talent contracts, union agreements and state right of publicity laws can restrict reuse of a person's voice or likeness. Counsel should review audio and video separately from text.
- Source confidentiality commitments
- Promises made to sources, and reporter's privilege protections that vary by state, cover unpublished notes and recordings. These materials should be treated as outside any license.
This page is general information, not legal advice. Whether your company can license a given dataset depends on your contracts, your privacy notices and the law where you and your customers are. Review it with your own counsel before signing anything.
Excluded
What we will not license.
- Wire service copy, syndicated columns and stock photography licensed in from others.
- Freelance work where the contributor agreement does not grant the needed rights and the contributor has not agreed.
- Unpublished reporting notes, raw interviews and anything that could identify a confidential source.
- Subscriber and reader personal data, including comment account details.
- Video or audio containing licensed music or third-party footage you cannot sublicense.
Data types common in media and publishers
Questions
Media and Publishers: what companies ask.
Can a publisher license its archive to AI labs?
Do we give up our copyright?
What happens to freelance pieces with unclear rights?
Can we limit how the content is used?
Find out if your data qualifies.
Eleven questions, about five minutes, no call required. You get an answer on screen and a written follow-up.