Getting started
Are business records copyrighted? What that means for licensing them
By SourceX Editorial · Reviewed by Noah Loul ·
Short answer
Business records are partly copyrighted: original writing such as emails, playbooks and code is usually protected, while the facts inside them, like dates, amounts and statuses, are not. Because coverage is patchy and authorship is mixed, data licenses rely mainly on contract terms, backed by trade secret protection, to define what a buyer may do.
Key takeaways
- Copyright generally protects original expression in records, not the facts, figures and statuses they contain.
- Work employees write within the scope of their jobs is generally owned by the employer as work made for hire; contractor work often is not.
- Customer messages, client deliverables and third-party documents inside your systems may belong to someone else.
- Trade secret protection depends on reasonable secrecy measures, so confidentiality terms in a license matter.
- Licenses set permitted use, exclusions and deletion in contract terms because copyright alone cannot.
Are business records copyrighted?#
Business records are copyrighted to the extent they contain original expression, and not for the facts they record. An engineer's written explanation of a design change, a support agent's troubleshooting reply or a carefully drafted SOP is likely protected expression. A ticket's timestamp, an order quantity and a job status are facts, which copyright does not protect.
Databases and exports sit in between. A compilation of facts may be protected for an original selection or arrangement, but that protection is thin and does not reach the underlying facts. A CRM export is therefore mostly unprotected data inside a structure that may or may not be original.
For licensing, copyright alone rarely tells you what a buyer may do with your records. Ownership, confidentiality and contract terms usually do more of the work.
Record types and the protection that usually applies#
Record types differ in how much copyright, trade secret and contract protection they carry, and most archives mix several types. The table is a general guide for a first conversation with counsel, not a legal conclusion about any specific record.
| Record type | Copyright | Trade secret | Contract terms that often control |
|---|---|---|---|
| Employee emails and chat | Often protected expression; usually company-owned if written as part of the job | Possible for pricing, strategy and methods | Employee confidentiality agreements and internal policies |
| Support tickets | Agent replies protected; customer messages may belong to customers | Rarely, apart from internal notes | Customer terms of service and privacy notices |
| CRM and transaction fields | Mostly unprotected facts | Possible for customer lists and pricing | Customer contracts and NDAs |
| SOPs, playbooks and wikis | Usually protected expression | Often, if kept confidential | Employee and contractor agreements |
| Source code and code reviews | Protected; open-source and third-party code carries its own licenses | Often, for proprietary code | Open-source licenses and customer contracts for custom work |
| Drawings and design documents | Protected; ownership set by project agreements | Sometimes | Owner-architect and owner-engineer agreements |
| Client deliverables | Protected, often owned by the client | The client's to claim, if anyone's | Master services agreements and statements of work |
| Vendor manuals and attachments | Owned by the vendor | Not yours to claim | Vendor terms of use |
- 01Your work records
- 02Personal details removed
- 03You approve the release
- 04Licensed to AI developers
Who owns copyright in what employees, contractors and customers wrote#
Copyright in writing by employees generally belongs to the employer when it is a work made for hire. Under 17 U.S.C. 201(b), the employer is considered the author of a work made for hire and owns the rights unless a signed written agreement says otherwise. Writing falls within the scope of employment when it is the kind of work the person was hired to do, done substantially within work time and space, and meant at least in part to serve the employer.
Contractors are different. A commissioned work counts as made for hire only if it falls in listed categories, such as a contribution to a collective work, a compilation or an instructional text, and the parties expressly agree in a signed writing. Code, reports and documents from consultants may therefore belong to the consultant unless an assignment clause moved them to you.
Customers wrote part of most support and CRM histories. Their messages, attachments and uploaded files may be theirs, subject to whatever your terms of service granted you. That is one reason licenses often keep only the company's side of an exchange, or remove customer content during preparation.
Trade secret protection often matters more than copyright#
Trade secret protection often matters more than copyright for business records, because it can cover facts, methods and compilations that copyright leaves open. Under the Defend Trade Secrets Act, information qualifies only if its owner has taken reasonable measures to keep it secret and it derives independent economic value from not being generally known.
Reasonable measures need not be absolute. Department of Justice guidance cites examples such as telling employees the information exists, limiting access to those who need it, requiring confidentiality agreements and keeping documents locked. The practical point for licensing: sharing records without confidentiality obligations can weaken the protection that makes them worth licensing.
That is why data licenses usually carry confidentiality, security and use restrictions even when records are de-identified. Whether a given disclosure preserves trade secret status is a question for counsel, assessed deal by deal.
What recent AI copyright rulings signal for record owners#
Recent AI copyright rulings signal that courts weigh training uses case by case, which makes documented permission valuable to buyers. None of the matters below involved a company licensing its own operating records, but they shape how model developers think about rights.
Commentators noted that the Ross case involved a non-generative tool built to compete with its source, leaving generative training questions to other courts. For a company licensing its own records, the practical lesson is that a written license removes guesswork for both sides.
| Ruling or report | What it said | Why record owners notice |
|---|---|---|
| Thomson Reuters v. Ross, Third Circuit, September 2026 | Affirmed that Westlaw headnotes are copyrightable and that copying them to train a legal-research AI was not fair use | Text written by a business's own staff can carry real copyright weight in AI training |
| Two June 2025 district court rulings on books | Found training language models on the books was fair use on those records; one left claims over pirated copies for trial | Outcomes turn on specific facts, so buyers value clear licenses |
| US Copyright Office pre-publication report on generative AI training, May 2025 | Covered when training may or may not be fair use and the licensing of works for AI training, and favored letting voluntary licensing markets develop | Licensing is part of the policy picture, not a side issue |
Why data licenses rely on contract terms#
Data licenses rely on contract terms because copyright and trade secret law leave gaps that only an agreement can close. A license binds the buyer to specific uses whether or not each record is protected, and it covers facts, metadata and de-identified content that copyright never reached.
Warranties deserve the closest attention from counsel. A supplier should promise only what its rights review supports, which is far easier when the review documented where each record family came from.
- Defined scope: the record families, systems and date ranges included, and what is excluded.
- Permitted use: training, evaluation or both, and any field-of-use limits.
- Confidentiality and security obligations that support trade secret protection.
- A ban on re-identification and on attempts to recover removed details.
- Retained ownership: the records are licensed, not sold, and the company keeps title.
- Third-party content: customer, client and vendor material removed or excluded.
- Term, deletion or return of delivered copies, and what survives the end of the license.
- Warranties limited to what the company actually knows about its rights.
Illustrative: an architecture and engineering firm separates drawings from decisions#
Illustrative: a fictional architecture and engineering firm considered licensing records from its project history: RFI responses in Procore, internal design review comments, Bluebeam markups and submittal logs. Its general counsel began by sorting the archive by who owned what.
Standard AIA owner-architect forms treat the architect and each of its consultants as the authors and owners of their respective instruments of service, and the firm's agreements followed that model. Drawings from its structural and MEP consultants therefore belonged to those consultants, and owner-provided surveys and reports came with confidentiality terms.
The firm scoped the license to its own RFI responses, review comments and submittal decisions, with project and client names removed, and excluded consultant drawings and owner documents entirely. The license rested on use restrictions and confidentiality terms rather than on copyright in any single page.
How SourceX handles rights in records#
SourceX handles ownership questions in the Rights step of the SourceX five-step transaction: Supply, Rights, Preparation, Approval and Delivery. Rights review looks at authorship, customer and vendor contracts, employee and contractor agreements and third-party content before any record is prepared.
Each license that proceeds carries a SourceX Evidence Packet, which records five things: provenance, licensing rights, permitted use, the privacy record and release authorization. On a rights question, the packet shows which record families were cleared and on what basis. The supplier keeps title, since records are licensed rather than sold, approves each step, and relies on its own counsel for legal conclusions.
Frequently asked questions
Do we need to register copyright before licensing records?
Generally no. A data license grants permission to use material under contract, and it does not depend on registration. Registration affects enforcement and remedies if a dispute ever arises, which counsel can explain. Few operating records are registered, and most licenses never rely on it.
Can we license records that contain customers' writing?
Possibly, depending on what your terms of service, contracts and privacy notices allow, and on the privacy laws that may apply. Many licenses remove customer-written content or keep only the company's side of the exchange. Counsel assesses this deal by deal, one record family at a time.
Does de-identification remove copyright concerns?
No. De-identification addresses privacy, not authorship. A de-identified support reply is the same written expression, owned by whoever owned it before. Preparation reduces privacy burden, while ownership and permission questions are settled during rights review.
Is our customer list protected by copyright?
Usually not in any meaningful way, because names and contact details are facts. A customer list is more often protected as a trade secret, if the company kept it confidential, and through contracts. That is why customer lists are normally excluded from licensed records altogether.
What about repositories that include open-source code?
Open-source code in your repositories carries its own license terms, which can set conditions on copying and redistribution. Code reviews, commit messages and issue discussions written by your engineers are different and usually company-owned. Rights review separates the two before any engineering records are scoped.
Sources
- 17 U.S.C. 201(b) provides that in the case of a work made for hire, the employer or other person for whom the work was prepared is considered the author and, unless the parties have expressly agreed otherwise in a written instrument signed by them, owns all of the rights comprised in the copyright. Source
- The Ninth Circuit model jury instruction states a work is within the scope of employment if it is the kind of work the employee is employed to perform, occurs substantially within authorized work hours and space limits, and is actuated at least in part by a purpose to serve the employer. Source
- 17 U.S.C. 101 defines a work made for hire as a work prepared by an employee within the scope of employment, or a specially ordered or commissioned work in listed categories, including a contribution to a collective work, a compilation or an instructional text, if the parties expressly agree in a signed written instrument. Source
- Under 18 U.S.C. 1839(3), information qualifies as a trade secret only if the owner has taken reasonable measures to keep it secret and it derives independent economic value from not being generally known. Source
- DOJ guidance states that trade secret protective measures need not be absolute but must be reasonable under the circumstances, citing advising employees of the secret's existence, need-to-know access, confidentiality agreements and locked documents. Source
- In late September 2026, a Third Circuit panel affirmed that the Westlaw headnotes are copyrightable and that ROSS's copying of them to train a legal-research AI was not fair use. Source
- The Ross case involved a non-generative AI tool, and commentators noted that the ruling left the questions of transformativeness and market effect in generative-AI training to other courts. Source
- On June 23, 2025, a judge of the Northern District of California held that using books to train large language models was exceedingly transformative and therefore fair use. Source
- In the same June 2025 ruling, downloading pirated books to build a permanent central library was held not to be fair use, and that infringement claim was left for trial. Source
- On June 25, 2025, a judge of the Northern District of California held that using the plaintiff authors' books to train language models was fair use on the record the plaintiffs presented. Source
- On May 9, 2025, the US Copyright Office released a pre-publication version of Copyright and Artificial Intelligence, Part 3: Generative AI Training, covering when training may or may not be fair use and the licensing of copyrighted works for AI training. Source
- The Copyright Office's Part 3 pre-publication report recommends letting voluntary licensing markets for AI training keep developing without government intervention. Source
- Under AIA B101-2017 Section 7.2, the Architect and the Architect's consultants are deemed the authors and owners of their respective Instruments of Service and reserve all rights, including copyrights. Source
Related resources
See if your company qualifies
A short company assessment. No data uploads are needed.