# How to sell photo and image libraries for AI training

URL: https://sellmydatatoai.net/guides/sell-photo-and-image-libraries-for-ai-training
Reviewed: 2026-11-08

You can license a photo or image library to AI companies where you hold the copyright in each image and releases for identifiable people. Shutterstock's chief executive told Bloomberg its AI licensing business produced $104 million in revenue in 2023, but no buyer publishes a per-image rate. Original files with captions and rights records are worth most. Commissioned photos, stock images and faces need checking first.

## Key takeaways

- Shutterstock's AI licensing business produced $104 million in revenue in 2023, according to its chief executive, speaking to Bloomberg.
- No buyer publishes a per-image rate. Getphyllo's reported ranges run from $5,000 to $15,000 for research licenses to $1 million and above for exclusive deals.
- Under US law a freelancer's commissioned photos belong to the photographer unless assigned in writing or covered by a qualifying work-for-hire agreement.
- Getty Images' license lists machine learning and AI use as restricted, so stock images usually have to come out.
- Screen every library for illegal content and strip GPS data from EXIF metadata before any sample leaves your company.

Licensed images are already a business for the largest libraries. Shutterstock's data licensing page claims more than 640 million images, and TechCrunch reported in July 2023 that OpenAI signed a six-year agreement to license Shutterstock images, videos, music and metadata. Smaller holders, from publishers to companies with years of product, site or inspection photos, can license too, on a smaller scale.

## Who licenses image libraries for AI training?

Image libraries are licensed through brokers, marketplaces and training-data companies, while the largest stock libraries license their own collections directly.

| Route | Example | What it publishes |
| --- | --- | --- |
| Data broker | Protege | Licenses healthcare, video, spatial and other domain data, with no partner terms published |
| Data marketplace | Defined.ai | Image datasets from partners who pass ethical vetting, 30% approved |
| Data marketplace | Opendatabay | Seller sets the price and keeps 70% to 95% |
| Rights-holder marketplace | Human Native | Content from creators and rights holders, acquired by Cloudflare per Techzine in January 2026 |
| Training-data company | Mercor | Site photos within construction deals of $115,000 to $1 million |
| Content licensing platform | Shutterstock | Licenses its own contributor library, not company archives |

See our profiles of [Opendatabay](https://sellmydatatoai.net/data-buyers/opendatabay) and [Shutterstock data licensing](https://sellmydatatoai.net/data-buyers/shutterstock-data-licensing) for fees and what each route suits.

## How much is an image library worth?

An image library has no published per-image price from any buyer we have checked. Getphyllo's reported market ranges, which are not quotes, give the clearest guide by license type.

| License type | Reported range |
| --- | --- |
| Research | $5,000 to $15,000 |
| Commercial | $25,000 to $500,000 and above |
| Professional, annual | $25,000 to $50,000 a year |
| Enterprise, annual | $50,000 to $500,000 a year |
| Exclusive or custom | $1 million and above |

Getphyllo notes that a research license rarely permits production training, so compare like with like. Our guide to [exclusive and non-exclusive licenses](https://sellmydatatoai.net/guides/exclusive-vs-non-exclusive-data-licenses) explains why exclusivity moves the price so far.

## What makes an image library more valuable?

An image library is more valuable when each image is an original with an accurate caption and a rights record, and when the subject matter is hard to find elsewhere.

| Factor | Worth more | Worth less |
| --- | --- | --- |
| Files | Original resolution, as captured | Compressed web copies |
| Captions | Written by people, specific and accurate | Missing or machine-written |
| Rights | A rights and release record for every image | Unknown photographer or source |
| Subject | Specialist, technical or inspection images | Generic scenes widely available |
| Labels | Expert labels or pass and fail decisions | No labels |
| Duplicates | Near-duplicates removed | Many crops of the same shot |

## Who owns the copyright in your photos?

Your company owns the copyright in photos taken by staff photographers as part of their jobs, which the US Copyright Office's Circular 30 treats as works made for hire. A freelancer's commissioned photos qualify only if they fit one of nine categories, such as a contribution to a collective work, and both sides signed a written agreement saying so. Otherwise the photographer owns them unless they assigned the copyright.

| Image source | Usual rights position | Usual approach |
| --- | --- | --- |
| Staff photographers and employees | Company-owned as works made for hire | Include |
| Commissioned freelancers | Photographer owns unless assigned or a qualifying work for hire | Include only with a written agreement |
| Stock images | Licensed, often with AI restrictions | Remove |
| Customer or user uploads | Governed by your terms with them | Check the terms, usually exclude |
| Images of artworks and branded products | Third-party rights in what is shown | Review or remove |

Stock licenses rarely allow training. Getty Images' content license agreement lists machine learning and AI use as restricted unless an additional license is bought, and bars use in connection with training, fine-tuning or other data ingestion.

## Do photos of people need releases?

Photos of identifiable people generally need model releases or another clear basis before you license them. New York's Civil Rights Law section 50 makes it a misdemeanor to use a living person's name, portrait or picture for advertising or trade without written consent, and whether a training license is trade is a question for your counsel.

Illinois' Biometric Information Privacy Act lists a scan of face geometry as a biometric identifier, while excluding photographs themselves. The risk arises when faces are processed, for example by a buyer's face analysis. Labelbox Prism requires explicit, timestamped consent for likeness data, which shows the standard buyers now expect. Leave out images of children.

## What must you screen before licensing images?

You must screen a library for illegal content and hidden personal data before licensing it. In December 2023 the Stanford Internet Observatory reported more than 3,200 images of suspected child sexual abuse in the LAION-5B dataset, which was taken down as a result, so hash-match every image against known illegal content.

- Strip GPS coordinates from EXIF metadata, which can reveal homes and sites
- Remove camera serial numbers and photographer names from metadata, keeping them in your rights record
- Blur or remove faces, license plates and screens where you lack releases
- Flag logos, artworks and trademarks in shot

## How should you package an image library?

You should package an image library as original files with a manifest that ties each image to its caption, labels and rights record.

- Original-resolution JPEG, PNG, TIFF or RAW files
- Captions and tags in JSON or Parquet, with the source of each caption marked
- Labels in a standard format such as COCO, with the labeling guide attached
- A rights record ID per image, and perceptual hashes for duplicate checks

Our page on [image datasets](https://sellmydatatoai.net/buy-ai-training-data/image-datasets) lists what AI teams check on delivery, and the guide to [data provenance](https://sellmydatatoai.net/guides/data-provenance-explained) explains the records buyers ask for.

## If you hold UK or EU data

Under section 11 of the UK's Copyright, Designs and Patents Act 1988, the photographer is the first owner of a commissioned photo unless they are your employee or assign it. The ICO says images used to uniquely identify a person are biometric data, a special category, and photos of identifiable people are personal data in any case.

## How do you start licensing an image library?

You start by counting images by source, year and subject, and recording which have captions, releases and a known photographer. That audit shows how much of the library is licensable before any buyer sees an image.

Every seller passes our eight-point vetting standard, including third-party content and a representative sample drawn at random. [Book a free consultation](https://sellmydatatoai.net/book-a-consultation?side=seller) to find out whether buyers would want your library.

## Frequently asked questions

### Can I license stock photos my company bought?

Stock photos your company bought are licensed to you under the library's terms, not owned by you, so you usually cannot license them for AI training. Getty Images' content license agreement, for example, lists machine learning and AI use as restricted unless an additional license is bought. Remove stock images and record their removal in the manifest.

### Do I own photos a freelancer took for my company?

Under US law you own a freelancer's photos only if they assigned the copyright to you in writing, or the photos fall in one of the nine work-for-hire categories, such as a contribution to a collective work, with a signed agreement saying so. Otherwise the photographer owns them. Check every commissioning agreement before including a shoot.

### Do images of employees need consent before licensing?

Images of identifiable employees should have a release or another clear basis before licensing. New York's Civil Rights Law section 50 restricts commercial use of a person's picture without written consent, and Labelbox Prism requires explicit, timestamped consent for likeness data. Where you lack consent, blur faces or leave the images out.

### What image metadata should I remove before licensing?

Remove GPS coordinates, camera serial numbers, photographer and editor names, and any client or project names from EXIF and IPTC metadata before licensing. Keep capture dates, camera model and lens if they are useful to the buyer. Hold the photographer and rights details in your own rights record, linked to each image by ID.

## Sources

- [Shutterstock's AI-licensing business generated $104 million](https://news.bloomberglaw.com/ip-law/shutterstocks-ai-licensing-business-generated-104-million), Bloomberg Law, accessed November 2026
- [Data licensing](https://www.shutterstock.com/data-licensing), Shutterstock, accessed November 2026
- [Shutterstock expands deal with OpenAI to build generative AI tools](https://techcrunch.com/2023/07/11/shutterstock-expands-deal-with-openai-to-build-generative-ai-tools), TechCrunch, accessed November 2026
- [Best public data providers for AI training and grounding](https://www.getphyllo.com/post/best-public-data-providers-ai-training-grounding), Getphyllo, accessed November 2026
- [Circular 30: works made for hire](https://www.copyright.gov/circs/circ30.pdf), US Copyright Office, accessed November 2026
- [Getty Images content license agreement](https://www.gettyimages.com/eula), Getty Images, accessed November 2026
- [New York Civil Rights Law section 50: right of privacy](https://www.nysenate.gov/legislation/laws/CVR/50), New York State Senate, accessed November 2026
- [Biometric Information Privacy Act, 740 ILCS 14/10 definitions](https://lig.ilga.gov/legislation/ilcs/documents/074000140K10.htm), Illinois General Assembly, accessed November 2026
- [Prism](https://prism.labelbox.com/), Labelbox, accessed November 2026
- [AI image dataset is pulled after child sex abuse pictures discovered](https://petapixel.com/2023/12/20/ai-image-dataset-is-pulled-after-child-sex-abuse-pictures-discovered), PetaPixel, accessed November 2026
- [Cloudflare acquires Human Native for AI data marketplace](https://www.techzine.eu/news/infrastructure/137991/cloudflare-acquires-human-native-for-ai-data-marketplace), Techzine, accessed November 2026
- [How to sell company data to AI companies (2026 guide)](https://ddiy.co/sell-company-data-to-ai/), ddiy.co, accessed November 2026
- [Copyright, Designs and Patents Act 1988, section 11](https://www.legislation.gov.uk/ukpga/1988/48/section/11), legislation.gov.uk, accessed November 2026
- [What is special category data?](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/lawful-basis/special-category-data/what-is-special-category-data/), Information Commissioner's Office, accessed November 2026

---

Book a free consultation: https://sellmydatatoai.net/book-a-consultation