Who takes 3D models for AI training
2 of 8 organisations read for this board take 3D models from the people who made them, read 27 October 2026.
How it acquires 3D models
Kind of data named
The data named, word for word
The first step for an owner
Ownership or consent, in its words
Every answer from every organisation
Each row is one organisation. Click an answer to see the sentence behind it, quoted from the page with the date we read it. Where it says "not stated", the page we read is silent on that question.
| Organisation | How it acquires 3D models | Kind of data named | The data named, word for word | The first step for an owner | Ownership or consent, in its words |
|---|---|---|---|---|---|
| A23D | sells a library it made itself | 3D models and assets | The dataset contains more than 579,000 assets, including 3D models, PBR materials… | not statedNot addressed in 3D training data for AI | This provides consistent quality, traceability, and clear dataset provenance across the… |
| CGAxis | sells a library it made itself | 3D models and assets | 7,794 PBR material sets covering brick, stone, concrete, wood, metal, fabric, and more. | not statedNot addressed in Licensed 3D training data for AI | We warrant that we own what we license, and we stand behind it. |
| Daz 3D | not statedNot addressed in Responsible 3D AI training data | 3D models and assets | Millions of structured 3D assets enabling infinite dataset generation | not statedNot addressed in Responsible 3D AI training data | Commercially safe training data with clear licensing for enterprise deployment |
| Defined.ai | runs a marketplace or intake where owners list data | many kinds, 3D among them | Scientific and technical content across science, technology, engineering, and mathematics. | Our in-house experts will guide you along the way in four easy steps. | Assessment of your licensing rights as well as legal, compliance and cybersecurity checks |
| InfoBay AI (AWS Marketplace) | not statedNot addressed in 1.3M+ 3D CAD models listing | 3D models and assets | The collection contains engineering-grade 3D models in industry-standard formats… | not statedNot addressed in 1.3M+ 3D CAD models listing | not statedNot addressed in 1.3M+ 3D CAD models listing |
| Modelry (CGTrader) | sells a library it made itself | 3D models and assets | Select your desired object categories, volume, licensing, and quality requirements. | not statedNot addressed in 3D model datasets for generative AI | Choose royalty-free assets for cost savings or full ownership for exclusivity. All… |
| Protege | licenses models from the people who made them | many kinds, 3D among them | Protege works with data and content providers across domains. | Select an industry below to learn more. | not statedNot addressed in For data providers |
| Shutterstock | licenses models from the people who made them | varies by document | TurboSquid 3D Models: Over 1 million models in formats like FBX, OBJ, USD, glTF, and more. | Tell us about your content library. We’ll follow up to discuss next steps.Not addressed in 3D and spatial AI datasets | All assets are rights-cleared and contributor-permissioned for AI training and… |
The 45 sentences the answers come from
- A23D, 3D training data for AI, read 2026-10-27, How it acquires 3D models:
Every asset is created by our in-house artists, with no AI-generated geometry or synthetic source content.
- A23D, 3D training data for AI, read 2026-10-27, Kind of data named:
The dataset contains more than 579,000 assets, including 3D models, PBR materials, decals, foliage, characters, and IES lighting profiles spanning over 625 categories.
- A23D, 3D training data for AI, read 2026-10-27, The data named, word for word:
The dataset contains more than 579,000 assets, including 3D models, PBR materials, decals, foliage, characters, and IES lighting profiles spanning over 625 categories.
- A23D, 3D training data for AI, read 2026-10-27, The first step for an owner: Not addressed: the page read on 2026-10-27 (4,782 characters) is silent on this.
- A23D, 3D training data for AI, read 2026-10-27, Ownership or consent, in its words:
This provides consistent quality, traceability, and clear dataset provenance across the entire corpus.
- CGAxis, Licensed 3D training data for AI, read 2026-10-27, How it acquires 3D models:
We have built our own 3D assets since 2008, in house, and we now license that catalogue as clean, single-owner, indemnified AI training data.
- CGAxis, Licensed 3D training data for AI, read 2026-10-27, Kind of data named:
4,390 3D models across furniture, decor, architecture, nature, and everyday objects.
- CGAxis, Licensed 3D training data for AI, read 2026-10-27, The data named, word for word:
7,794 PBR material sets covering brick, stone, concrete, wood, metal, fabric, and more.
- CGAxis, Licensed 3D training data for AI, read 2026-10-27, The first step for an owner: Not addressed: the page read on 2026-10-27 (6,873 characters) is silent on this.
- CGAxis, Licensed 3D training data for AI, read 2026-10-27, Ownership or consent, in its words:
We warrant that we own what we license, and we stand behind it.
- Daz 3D, Responsible 3D AI training data, read 2026-10-27, How it acquires 3D models: Not addressed: the page read on 2026-10-27 (4,283 characters) is silent on this.
- Daz 3D, Responsible 3D AI training data, read 2026-10-27, Kind of data named:
Train AI models faster using fully licensed, structured 3D datasets designed for scalable, reliable machine learning.
- Daz 3D, Responsible 3D AI training data, read 2026-10-27, The data named, word for word:
Millions of structured 3D assets enabling infinite dataset generation
- Daz 3D, Responsible 3D AI training data, read 2026-10-27, The first step for an owner: Not addressed: the page read on 2026-10-27 (4,283 characters) is silent on this.
- Daz 3D, Responsible 3D AI training data, read 2026-10-27, Ownership or consent, in its words:
Commercially safe training data with clear licensing for enterprise deployment
- Defined.ai, Partnership programs, read 2026-10-08, How it acquires 3D models:
Monetize your data ethically and easily with Defined.ai’s Partnership Programs.
- Defined.ai, Partnership programs, read 2026-10-08, Kind of data named:
Audio, video, image, text or multimodal data
- Defined.ai, Partnership programs, read 2026-10-08, The data named, word for word:
Scientific and technical content across science, technology, engineering, and mathematics.
- Defined.ai, Partnership programs, read 2026-10-08, The first step for an owner:
Our in-house experts will guide you along the way in four easy steps.
- Defined.ai, Partnership programs, read 2026-10-08, Ownership or consent, in its words:
Assessment of your licensing rights as well as legal, compliance and cybersecurity checks
- InfoBay AI (AWS Marketplace), 1.3M+ 3D CAD models listing, read 2026-10-27, How it acquires 3D models: Not addressed: the page read on 2026-10-27 (12,443 characters) is silent on this.
- InfoBay AI (AWS Marketplace), 1.3M+ 3D CAD models listing, read 2026-10-27, Kind of data named:
This dataset is a large-scale collection of 3D CAD models designed to support engineering AI
- InfoBay AI (AWS Marketplace), 1.3M+ 3D CAD models listing, read 2026-10-27, The data named, word for word:
The collection contains engineering-grade 3D models in industry-standard formats including SolidWorks (SLDPRT), STEP, and GLB
- InfoBay AI (AWS Marketplace), 1.3M+ 3D CAD models listing, read 2026-10-27, The first step for an owner: Not addressed: the page read on 2026-10-27 (12,443 characters) is silent on this.
- InfoBay AI (AWS Marketplace), 1.3M+ 3D CAD models listing, read 2026-10-27, Ownership or consent, in its words: Not addressed: the page read on 2026-10-27 (12,443 characters) is silent on this.
- Modelry (CGTrader), 3D model datasets for generative AI, read 2026-10-27, How it acquires 3D models:
Upon approval, we create the assets according to your specifications.
- Modelry (CGTrader), 3D model datasets for generative AI, read 2026-10-27, Kind of data named:
Custom 3D datasets tailored to your exact technical specifications, complete with full licensing for Generative AI applications.
- Modelry (CGTrader), 3D model datasets for generative AI, read 2026-10-27, The data named, word for word:
Select your desired object categories, volume, licensing, and quality requirements.
- Modelry (CGTrader), 3D model datasets for generative AI, read 2026-10-27, The first step for an owner: Not addressed: the page read on 2026-10-27 (2,625 characters) is silent on this.
- Modelry (CGTrader), 3D model datasets for generative AI, read 2026-10-27, Ownership or consent, in its words:
Choose royalty-free assets for cost savings or full ownership for exclusivity. All licenses include AI usage rights
- Protege, For data providers, read 2026-10-08, How it acquires 3D models:
Protege works with data and content providers across domains. Select an industry below to learn more.
- Protege, For data providers, read 2026-10-08, Kind of data named:
Protege works with data and content providers across domains.
- Protege, For data providers, read 2026-10-08, The data named, word for word:
Protege works with data and content providers across domains.
- Protege, For data providers, read 2026-10-08, The first step for an owner:
Select an industry below to learn more.
- Protege, For data providers, read 2026-10-08, Ownership or consent, in its words: Not addressed: the page read on 2026-10-08 (21,248 characters) is silent on this.
- Shutterstock, 3D and spatial AI datasets, read 2026-10-27, How it acquires 3D models:
All assets are rights-cleared and contributor-permissioned for AI training and commercialization.
- Shutterstock, 3D and spatial AI datasets, read 2026-10-27, Kind of data named:
one of the largest licensed collections of 3D models and photorealistic renders
- Shutterstock, 3D and spatial AI datasets, read 2026-10-27, The data named, word for word:
TurboSquid 3D Models: Over 1 million models in formats like FBX, OBJ, USD, glTF, and more.
- Shutterstock, 3D and spatial AI datasets, read 2026-10-27, The first step for an owner: Not addressed: the page read on 2026-10-27 (4,650 characters) is silent on this.
- Shutterstock, 3D and spatial AI datasets, read 2026-10-27, Ownership or consent, in its words:
All assets are rights-cleared and contributor-permissioned for AI training and commercialization.
- Shutterstock, Data Partner Network, read 2026-10-08, How it acquires 3D models:
Explore Shutterstock’s Data Partner Network to license your content for AI training
- Shutterstock, Data Partner Network, read 2026-10-08, Kind of data named:
What type of content do you control (select all that apply)
- Shutterstock, Data Partner Network, read 2026-10-08, The data named, word for word:
What type of content do you control (select all that apply)
- Shutterstock, Data Partner Network, read 2026-10-08, The first step for an owner:
Tell us about your content library. We’ll follow up to discuss next steps.
- Shutterstock, Data Partner Network, read 2026-10-08, Ownership or consent, in its words:
Retain ownership and control
Documents we read but did not count
After reading, 2 named documents were excluded as off the board's subject; the reasons sit below. They count as read but give no answers.
- Wirestock, Earning from AI training datasets as a creator: Wirestock's page names photos, videos and illustrations; it does not name 3D models among the content it licenses
- Troveo, Data for frontier labs: Troveo's page names video, audio, gaming and company data; it does not name 3D models among what it licenses
The method in full
Each organisation's own data licensing and partner pages were read through estate-fetch and stored whole (sha256 on every row).
For every page, five questions were answered from its own words: how it acquires 3D models, kind of data named, the data named, word for word, the first step for an owner, ownership or consent, in its words.
Every stated answer carries the sentence it rests on, checked against the stored text; 'not stated' means the page read does not address the question.
A organisation with several pages reads 'varies' where they disagree.
8 organisations, 9 pages.
11 sources, with links
- Shutterstock, 3D and spatial AI datasets (from the publisher's own data licensing and partner pages; reproduced for comment)
- Shutterstock, Data Partner Network (from the publisher's own data licensing and partner pages; reproduced for comment)
- CGAxis, Licensed 3D training data for AI (from the publisher's own data licensing and partner pages; reproduced for comment)
- Protege, For data providers (from the publisher's own data licensing and partner pages; reproduced for comment)
- Defined.ai, Partnership programs (from the publisher's own data licensing and partner pages; reproduced for comment)
- AWS Marketplace, 1.3M+ 3D CAD models listing (from the publisher's own data licensing and partner pages; reproduced for comment)
- Wirestock, Earning from AI training datasets as a creator (read, set aside)
- Troveo, Data for frontier labs (read, set aside)
- A23D, 3D training data for AI (from the publisher's own data licensing and partner pages; reproduced for comment)
- Daz 3D, Responsible 3D AI training data (from the publisher's own data licensing and partner pages; reproduced for comment)
- Modelry (CGTrader), 3D model datasets for generative AI (from the publisher's own data licensing and partner pages; reproduced for comment)
Use this figure
2 of 8 organisations read for this board take 3D models from the people who made them, read 27 October 2026.
How it acquires 3D models, by organisation: sells a library it made itself 3 of 8, licenses models from the people who made them 2 of 8, not stated 2 of 8, runs a marketplace or intake where owners list data 1 of 8 (8 organisations, read 27 October 2026).
Kind of data named, by organisation: 3D models and assets 5 of 8, many kinds, 3D among them 2 of 8, varies 1 of 8 (8 organisations, read 27 October 2026).
The data named, word for word, by organisation: stated 8 of 8 (8 organisations, read 27 October 2026).
The first step for an owner, by organisation: not stated 5 of 8, stated 3 of 8 (8 organisations, read 27 October 2026).
Ownership or consent, in its words, by organisation: stated 6 of 8, not stated 2 of 8 (8 organisations, read 27 October 2026).
Suggested citation: "Who takes 3D models for AI training", updated 2026-10-27, https://sell3dmodelstoai.com/datasets/3d-model-data-buyers/.
Download as CSV (45 answers; derived 2026-10-27)