Is business data copyrightable? Facts, compilations and originality

Why facts are not protected by copyright#

Facts are not protected by copyright because copyright covers original expression, not the information that expression describes. A delivery date, a part count or a temperature reading exists independently of whoever recorded it, so nobody can own it by writing it down.

US law does not reward effort on its own. A company that spent years building a complete record of every job, order or inspection does not gain copyright in the facts through that effort. What it may own is the way it expressed or organized them, which is a much narrower thing.

That does not make the records free to take. Trade secret law, confidentiality obligations and contracts can protect facts that copyright leaves open, which is why a data license never depends on copyright alone.

What does compilation copyright protect in a database?#

Compilation copyright protects the original selection, coordination or arrangement of material in a database, not the material inside it. If your team made creative choices about which fields to capture or how to classify records, those choices may be protected; the underlying entries are not.

The protection is often described as thin. A competitor that extracts the facts and arranges them differently may not infringe at all. Obvious arrangements, such as alphabetical or date order, or a schema any company in the industry would use, add little.

The United States has no separate database right of the kind the European Union grants for investment in databases. US owners of large operational databases therefore lean more on contracts and access controls than on intellectual property law.

Which business records contain copyrightable expression?#

Business records contain copyrightable expression wherever a person wrote, drew or coded something with some creative choice. The table sorts common operational records by how much expression they hold and who is likely to own it.

Who owns the expression in records your company holds?#

The expression in your records belongs to whoever created it, unless an employment relationship or a written agreement moved ownership. Holding a record is not the same as owning the copyright in it, and a rights review has to separate the two.

Text generated entirely by software raises a further question. US copyright generally requires human authorship, so automated status messages and machine-written summaries may carry little or no protection, even when your company owns the system that produced them.

• 1Employees: work prepared within the scope of their jobs is a work made for hire, so the employer is treated as the author and owner unless a signed agreement says otherwise.
• 2Contractors: their work usually stays theirs unless a signed agreement assigns it. Commissioned work counts as work made for hire only in certain statutory categories and only with a signed written agreement, so contractor code and reports generally need an assignment.
• 3Customers: text, files and code they send through support channels are typically their own, subject to your terms of service.
• 4Clients of professional firms: deliverables may be assigned to the client or licensed back under the engagement contract.
• 5Software vendors: templates and automated content in a SaaS platform may carry the vendor's terms.

Why contract matters more than copyright in a data license#

Contract matters more than copyright in a data license because the license binds the buyer whatever the copyright position of each record. Permitted use, confidentiality, no redistribution, deletion at term end and audit rights apply to facts and expression alike.

Copyright still matters in two places. It strengthens your remedies if records are misused, since an infringement claim can sit alongside breach of contract. And it limits what you can license: expression owned by customers, clients or contractors needs their permission or has to come out.

Courts hearing AI training disputes have looked closely at originality. In Thomson Reuters v. Ross, a federal court in Delaware held in February 2025 that Westlaw headnotes were original enough to protect and rejected Ross's fair-use defense for copying them to build a competing AI legal research tool. The Third Circuit affirmed in late September 2026, in what was reported as the first federal appellate ruling on fair use in AI training. The case turned on editorial writing, not raw facts, and involved a non-generative tool, so owners still write the important protections into the license.

Copyright assumptions that mislead data owners#

Copyright assumptions mislead data owners most often when they treat holding records as owning everything in them. Each of the beliefs below sounds reasonable and fails in a rights review.

Assumptions like these are cheapest to correct during the rights review, before a buyer's counsel raises them in a negotiation or a warranty schedule.

• 1We built the database, so we own everything in it: ownership of the compilation does not reach the facts inside it or text other people wrote.
• 2The records sit in our systems, so we can license them: customer, client and contractor content may need permission or removal.
• 3Facts are unprotected, so our data is open to anyone: contracts and trade secret law still restrict people who agree to terms or obtain records improperly.
• 4A copyright notice on the export protects it: a notice identifies the owner but does not create protection for facts.
• 5Our machine logs are protected because we generated them: purely automated output may have no human author at all.

Illustrative: a civil engineering firm sorts its project archive#

Illustrative: a fictional civil engineering firm wants to license records from its project archive. It holds project setup and hours in Deltek, marked-up plan sets in Bluebeam, RFI and submittal logs, internal design review comments, and final reports delivered to municipal and private clients.

Counsel sorts the archive by copyright position. Hours, dates and RFI status fields are facts. Internal review comments and RFI responses written by the firm's engineers are its own expression. Final reports and stamped drawings are client deliverables, and several client contracts assign ownership to the client.

The firm licenses the RFI workflow and internal review history, with client names and project addresses removed, and excludes the deliverables entirely. The license, not copyright, carries the use limits on the factual fields.

How SourceX treats copyright in the rights review#

SourceX treats authorship as one input to Rights, the second step of the SourceX five-step transaction: Supply, Rights, Preparation, Approval and Delivery. For each record family, the review asks who created the content, whether ownership was assigned and whether customer, client or vendor terms limit licensing.

Those answers land in the licensing rights section of the SourceX Evidence Packet. A buyer reading it can tell which records rest on owned expression, which are factual records licensed by contract, and which third-party content was taken out before delivery.