Short answer
A business that holds OEM service manuals usually has a right to use them for service work, not a right to license them for AI training. Copyright and the dealer or subscription agreement typically sit with the manufacturer. Buyers should ask who owns the content, what the access agreement permits, and whether the OEM will grant written permission before any license is signed.
This page raises diligence questions and clause topics. It is not legal advice; have counsel review any actual agreement.
The buyer's problem
Service companies, dealers and fleet owners hold large libraries of repair documentation, and they are often willing sellers. The difficulty is that holding a manual is not owning it. A repair shop that subscribes to an OEM's service portal accepted terms that usually limit use to servicing the OEM's products. A dealer received documents under a dealer agreement that may prohibit copying or sharing. Neither can pass on rights it does not have.
A buyer who signs with the holder alone can end up with a dataset it cannot use, or must delete. The check has to happen before money moves. For the content itself, see service manuals for embodied AI.
Who can grant what
| Party | Typically holds | Typically can license for training? | What to ask for |
|---|---|---|---|
| OEM | Copyright in manuals, figures, data modules | Yes, if it chooses | Written grant covering training and evaluation |
| Authorized dealer | Access under a dealer agreement | Rarely, without OEM consent | The agreement terms; OEM consent |
| Independent repair shop | Subscription or legacy copies | Rarely | Subscription terms; proof of any OEM permission |
| Fleet or facility owner | Manuals supplied with equipment | Depends on the supply contract | The purchase or service contract |
| Technical publisher/aggregator | Compiled or licensed-in content | Only for what it owns | Chain of title for each source |
| Technician | Own notes, photos, checklists | Yes, for original material, subject to employer terms | Employment and IP terms |
The last row matters. A shop's own work orders, technician notes and photographs may be licensable even when the OEM manual is not. See work order histories and repair shop networks as suppliers for what those assets offer.
What the data can support - and what it cannot
Documentation that is cleanly licensed supports procedure grounding, retrieval and task planning for the equipment it covers. It does not substitute for demonstrations of the physical work. And a license that covers reading the manual does not necessarily cover copying it into a training corpus, creating derived datasets, or letting a model reproduce passages. Those are separate permissions.
The U.S. Copyright Office's pre-publication report on generative AI training identifies stages of model development where copying of works may implicate rights, analyzes fair use, and examines licensing options (U.S. Copyright Office, 2025). It is useful for framing the questions to ask; it does not tell you whether a given use of a given manual is permitted. Check the contract terms with counsel rather than relying on general commentary.
Clause topics to cover
- Grant scope: training, fine-tuning, evaluation, retrieval, internal use; whether outputs may quote the content.
- Derived data: parsed text, chunks, embeddings, extracted tables and diagram annotations; who owns them.
- Redistribution: whether you may share the dataset or a model trained on it with customers or affiliates. See model weights, derivatives and redistribution.
- Exclusivity and field of use: whether the OEM may license the same material to competitors, and whether use is limited to a product line or purpose. See exclusivity and field of use.
- Warranties: the supplier's statement that it has the right to license, and what happens if that proves false.
- Revision and withdrawal: what happens when the OEM updates, recalls or revokes documents; whether you must delete superseded versions.
- Retention and audit: how long you may keep the data and what proof of deletion is required. See retention, deletion and audit rights.
- Format and ownership of structured sources: modules in a publishing standard may be owned by a prime contractor or customer; see S1000D and ATA documentation and who owns a procedure.
- Embedded third-party material: licensed fonts, supplier drawings, standards text and photographs inside manuals. See third-party content in recordings.
- Safety content: whether safety-critical instructions may be reproduced or must be handled with disclaimers.
Delivery mechanics
Delivery is part of the rights question. Documents obtained through a portal may include watermarks, account identifiers or access-controlled content that should not be transferred. Confirm what is delivered (files, structured exports, or access to a repository), whether identifiers are stripped, and whether the supplier can show the export did not breach its own access terms. Keep a record of the source version, date and permission for each document set, alongside a datasheet.
Illustrative permission checklist
# Illustrative only - synthetic values
document_set: oem_x_service_manuals_family_A
holder: repair_service_company_017
owner_of_content: oem_x
holder_access_basis: subscription_terms (copy on file)
oem_written_permission:
training: required, status = requested
evaluation: required, status = requested
derived_data: required, status = requested
embedded_third_party_content: reviewed, 6 items flagged
excluded: customer repair histories, pricing, dealer-only bulletins
restrictions_noted: no redistribution; deletion on request
Questions to ask suppliers
- Who is the copyright owner of each document set, and can you show the chain of title?
- What agreement gives the holder access, and does it prohibit copying, sharing or use for machine learning?
- Has the OEM been asked to permit training use, and is the answer in writing?
- Does the supplier also hold its own material (technician notes, checklists, photographs) that is separately licensable?
- Are any documents marked confidential, export-controlled or safety-critical?
- Which derived forms (parsed text, embeddings, annotations) does the license cover?
- How are revised or withdrawn documents handled after delivery?
- What happens to the agreement if the OEM objects after signing?
If the supplier cannot answer the first three, treat the offer as unverified.
Discuss your requirements
If you want repair documentation licensed for training, describe the equipment families and permitted uses to SourceX. SourceX looks for US businesses that hold the material you describe, checks the data and the supplier's licensing permissions, agrees pricing and allowed uses in a license, and coordinates delivery; nothing is contracted until a supplier agrees, and a request does not guarantee a match. For broader terms, see licensing terms for robotics training data, the humanoid training data guide and the maintenance and repair data hub. The site's licensing terms and data license terms explained cover license structure in general.
Explore sourcing and licensing options with SourceX
Sources
- Copyright and Artificial Intelligence, Part 3: Generative AI Training — framing of copying, fair use and licensing options in AI training.