AI training data and image licensing

After Getty and Shutterstock: who sells licensed image data in 2027

Three watches with pale numbered dials, gold-tone cases and contrasting bracelets and straps.
Three gold-tone wristwatches.Photograph by Robert Harper. Robert Harper Photograph Library, Archive.London. From the archive, shown for illustration.

The $3.7 billion merger is off and Getty is restructuring. For buyers of licensed image data, the supply side going into 2027 is more fragmented than anyone expected a year ago.

On 7 July 2026 the Competition and Markets Authority announced that Getty Images had walked away from its $3.7 billion merger with Shutterstock. The regulator had cleared the deal on one condition: that Shutterstock's editorial business, which the CMA's notice describes as global and as including the Backgrid and Splash brands, be sold to an approved buyer. Getty chose to terminate instead. Fifteen weeks later, on 21 October 2026, Getty filed a prepackaged Chapter 11 in Delaware.

The company that many expected to become the dominant single source of licensed visual data is instead two independent companies, one of them mid-restructuring. This article maps who is selling licensed image data as 2027 approaches. It reports what has happened and does not predict what will. Figures from third parties carry their source and date.

This article is general information, not legal advice. Checked on 18 November 2026.

What happened to the Getty and Shutterstock merger?

The CMA's account is short. Its interim report in February 2026 raised provisional concerns about editorial content supplied to UK media. After that report the companies proposed selling only Backgrid and Splash, which the CMA judged insufficient. Its final report in May 2026 allowed the merger only if the whole of Shutterstock's editorial business was divested. It found no competition concern in stock content. Getty then gave notice terminating the agreement, which the CMA described as a commercial choice.

Getty's Chapter 11 followed in October. According to Octus, reporting on the day of the filing, the prepackaged plan includes an equity investment of up to $533 million from Getty family members and affiliates of Koch Equity Development, and the company expects to finish by year-end as a privately held business. A disclosure statement filed with the US Securities and Exchange Commission says the company intended to ask for a confirmation hearing on 14 December 2026; we could not confirm on the date above that the plan had been confirmed. Getty said it would continue operating in the ordinary course, with no anticipated impact on contributors, partners or customers.

Where do Getty and Shutterstock stand on AI data now?

Both still sell data, on different terms.

Shutterstock has the more developed data business. Its data-licensing page cites 640M+ images, 70M+ videos and 3M+ contributors, and invites buyers to "Talk to our AI data team". Reuters reported on 5 April 2024 that Meta, Google, Amazon and Apple had each struck agreements with it, with initial deals worth $25 million to $50 million each according to its chief financial officer.

Getty Images has drawn a sharper line. It has a data-licensing page offering curated datasets, but chief executive Craig Peters told investors on 16 March 2026: "We do not license out our editorial content in any way, shape, or form with respect to AI training." Its agreement with OpenAI, announced on 21 June 2026, was reported as display-only: editorial images are retrieved and shown with credit in ChatGPT search, and are not used to train. Management said on the same March call that it expects traditional data licensing to be a low single-digit percentage of revenue.

One caution on the wider trend. Headline figures in this market come from a small number of large agreements, and Getty's own expectation that data licensing will stay a small share of its revenue is a reason to be wary of any claim that demand is growing in a straight line.

Who else sells licensed image data?

Supplier Type What is reported Source and date
Wirestock Broker and aggregator $23m Series A; 700,000+ contributors; supplies six large foundation-model makers; about $40m annual run-rate revenue TechCrunch, 14 May 2026
Defined.ai Dataset marketplace Sells datasets built from Getty Images content on its marketplace Defined.ai, 8 January 2025
Human Native (Cloudflare) Marketplace UK marketplace launched in 2024; acquisition by Cloudflare announced, terms undisclosed IT Brief, 16 January 2026
Protege Data platform $30m from Andreessen Horowitz; revenue share to data partners; holdings reported as video, audio and medical data SiliconANGLE, 8 January 2026
Bria Model developer that buys Training data from commercial agreements with 30+ data partners; attribution-based payment Bria's website, read 18 November 2026
Alamy Stock library Said in April 2023 it would license images for AI training for a fee, in non-image-generation uses, sharing revenue with contributors Alamy blog, April 2023
Dataset Providers Alliance Trade body Formed June 2024 with seven members, including Datarade, vAIsual and Pixta AI CIO, June 2024

Three observations follow from the table.

First, the fastest-growing intermediaries are built on contributor networks, not archives. Wirestock's model is creator-led, and whether such platforms take in historical collections is something a rights holder has to ask.

Second, not every "AI data marketplace" handles still images. Protege's reported holdings are video, audio, clinical records and medical imaging. Check the media types before assuming a fit.

Third, some of the demand side is visible. Bria publishes the fact that it buys, and lists a route to become a data partner. Fairly Trained, a certification body, lists 19 certified entries, most of them music and audio companies; its certificate attaches to model makers, so its list is better read as a guide to who cares about licensed inputs than as a directory of sellers.

What about public and collective schemes?

In the UK the government's Creative Content Exchange is a pilot marketplace for digitised cultural content. The Museums Association reported on 3 February 2026 that twelve bodies were taking part, among them the V&A, the National Portrait Gallery, the Natural History Museum, Imperial War Museums and The National Archives, with an operational pilot expected that summer. We could not confirm from a primary source whether the platform is live, and no route for private owners to join has been published.

Collective licensing for images has not arrived. The UK scheme being developed by the Copyright Licensing Agency with publishers' and authors' societies covers text. We found no launched collective licence for AI training on photographs.

Where do independent archives fit?

Among heritage picture libraries, AI licensing is mostly handled by saying no unless asked. Bridgeman Images' published restrictions bar machine-learning and AI use of its content and metadata unless specifically authorised in the licence. When we looked on 17 November 2026, the homepages of Bridgeman, TopFoto and Magnum Photos did not mention AI or data licensing; the dedicated pages we found belong to the large stock platforms. That is an observation about three homepages on one day, not a statement of any library's policy.

That leaves a gap between the bulk libraries and the closed door. A buyer who needs a specific domain, a documented single source, or a period of sole access will not find it in a 640-million-image catalogue licensed to all comers. A rights holder with a coherent collection and one signature can offer it directly, without a contributor agreement or an aggregator in between. The price of that route is scale: as our piece on fine-tuning versus foundation training sets out, collections in the tens of thousands suit fine-tuning, evaluation and specialist subjects, not pretraining.

How to compare suppliers

Whichever route a buyer takes, the questions are the same, and they are listed in what AI developers look for in a licensed image dataset. The short version:

  • How many contracts stand between you and the person who made the image?
  • Does the supplier's agreement with its contributors clearly cover the use you intend?
  • What documentation comes with the data, and what does it admit?
  • Is exclusivity possible, or is the same set on sale to everyone?
  • What has been reviewed and excluded, and by whom?

How the resulting agreements are priced and drafted is covered in how AI image licensing deals are structured.

Where Archive.London sits on this map

The Robert Harper Photograph Library is an independent, single-photographer archive of 95,487 unique photographs by inventory count, owned by 825 Ventures Ltd, the owner and rights holder. It covers about thirty-five years of editorial and commercial work, from studio still life to musicians and bands. It is small by the standards of this market, more than half of it is still on film, and it is presented for what it is: a single source, documented at collection level, whose size points to fine-tuning, evaluation and specialist domains. No price is published and no clearance claim is made for any image. Licences are negotiated directly with the single rights holder, subject to rights review.

Data buyers and brokers who want the dataset card, a sample set under NDA, or a first-look conversation as scanning proceeds can find the details on the AI training data licensing page.

Frequently asked questions

Did Getty Images and Shutterstock merge?

No. The $3.7 billion merger announced in January 2025 was terminated by Getty. The UK Competition and Markets Authority said on 7 July 2026 that it had cleared the deal on condition that Shutterstock's editorial business, which its notice describes as global and as including the Backgrid and Splash brands, was sold to an approved buyer, and that Getty had instead given notice ending the agreement. The companies remain independent.

Is Getty Images still operating after its Chapter 11 filing?

Yes, by its own account. Getty filed a prepackaged Chapter 11 in the US Bankruptcy Court for the District of Delaware on 21 October 2026. It said it would continue operating in the ordinary course, expected no impact on contributors, customers or partners, and aimed to complete the restructuring by the end of 2026, emerging as a privately held company. Buyers should check the current position.

Where can AI developers buy licensed image data?

From four kinds of supplier: large stock libraries with data-licensing teams, such as Shutterstock and Getty Images; brokers and aggregators such as Wirestock and Defined.ai; listing marketplaces such as Datarade; and rights holders licensing directly, including independent archives. Each differs in scale, documentation, exclusivity and how many contracts stand between the buyer and the original creator.

What is the Creative Content Exchange?

It is a UK government pilot marketplace through which AI developers could license digitised content from public cultural bodies. Twelve participants were named in early 2026, including the V&A, the National Portrait Gallery, Imperial War Museums and The National Archives. An operational pilot was expected in summer 2026; we could not confirm from primary sources whether it is live, and no joining route for private owners has been published.

Sources

This article is general information, not legal advice. It reflects the position as we understood it on the date shown and may be out of date. Published by Archive.London. Corrections are welcome under our editorial policy.

Licensing the archive for AI and data projects

95,487 photographs from one photographer and one rights holder. Read the dataset card, search the 503-folder inventory, discuss a sample set under NDA, or arrange a first look as scanning proceeds.

See AI and data licensingWrite to the archive