Proprietary data is information a company created or collected itself, controls, and does not make public: operating records, customer and transaction histories, designs, source code, pricing and the know-how written into procedures and tickets. It is proprietary because the company holds the rights and restricts access, not because the law gives it a special label. Some of it is also a trade secret, which US law protects only when the owner takes reasonable measures to keep it secret. Because it cannot be found on the public web, proprietary operational data is what AI developers license from companies.
- Proprietary data is owned or controlled by one company and kept from the public; proprietary information is the wider term for the same idea and covers know-how as well as datasets.
- A trade secret is the legally protected subset: secret, valuable because it is secret, and guarded by reasonable measures.
- Operational data, the records a business produces by running, is the largest share of most companies' proprietary data.
- Facts themselves are not protected by copyright, so contracts and access controls do most of the protecting.
- A company can license proprietary data without giving it away: the buyer gets a defined right to use a prepared copy.
What is proprietary data?
Proprietary data is data that belongs to one organization, is kept from the public and gives that organization an advantage because others cannot get it. Most of it is produced as a by-product of doing business: every order, ticket, inspection, claim, design revision and customer call leaves a record that exists nowhere else.
Proprietary information is the broader phrase, and the one that appears in contracts and government rules. The US government security glossary defines it as "material and information relating to or associated with a company's products, business, or activities", including financial information, data or statements, trade secrets, product research and development, designs, marketing plans, client lists, computer programs, processes and know-how that the company has clearly identified and marked as proprietary (NIST, from CNSSI 4009-2022). That definition adds a condition worth copying: the information must have been developed by the company and not be available without restriction from another source.
In practice, three things make data proprietary: the company created or lawfully collected it, the company controls who can see it, and it is not available elsewhere on the same terms.
Source: NIST CSRC glossary: proprietary information (from CNSSI 4009-2022)
Proprietary information vs confidential information vs trade secrets
Proprietary describes ownership and control; confidential describes a duty not to disclose; a trade secret is a legal status that only some proprietary information reaches. The same customer list can be all three, or only the first.
- Because facts are not copyrightable, the main protections for business records are secrecy, access control and contracts. That is why a data license spells out who may use the data and for what.
- Losing secrecy can end trade secret protection. Share proprietary data only under written terms and only the part that is needed.
| Term | What it means | Where the protection comes from |
|---|---|---|
| Proprietary data or information | Owned or controlled by the company and not public | Ownership, access control and contracts |
| Confidential information | Information someone has agreed, or is required, not to disclose | The confidentiality clause or non-disclosure agreement that defines it, and the duties of employees and advisers |
| Trade secret | Information that is kept secret by reasonable measures and "derives independent economic value, actual or potential, from not being generally known" (18 U.S.C. § 1839) | Trade secret law, such as the Defend Trade Secrets Act, plus the owner's own safeguards |
| Proprietary business information held by the US government | Company information shared with an agency is controlled as CUI in the category General Proprietary Business Information, marked PROPIN (National Archives CUI Registry) | Federal rules on controlled unclassified information |
| Copyright in a dataset | Only original selection, arrangement or expression. The Supreme Court held that "facts are not copyrightable" (Feist v. Rural) | Copyright law, which rarely covers raw operating records |
Source: 18 U.S.C. § 1839 (Defend Trade Secrets Act definitions), LII
Proprietary data examples by industry
Examples of proprietary data include work orders, claim files, maintenance records, underwriting decisions, call recordings, design files, project records and source code, as long as the company created them and keeps them private. The table lists typical examples in record-rich industries, with the reason an outsider cannot rebuild them.
- Not every item is licensable. Records that belong to a client, such as a carrier's claim files handled by a TPA, are the client's proprietary data, not yours.
- To list your own categories, use the data inventory template or map one workflow.
| Industry | Proprietary data examples | Why outsiders cannot rebuild it |
|---|---|---|
| Manufacturing | Work orders, travelers, nonconformance reports, inspection logs, machine maintenance histories | Years of production decisions tied to measured results on one shop floor |
| Insurance claims | Claim files with adjuster notes, reserves, photos, estimates and payments (see insurance claims data) | Each file records how a claim was decided and what it finally cost |
| Aviation maintenance | Task cards, logbook entries, defect and parts histories (see aircraft maintenance records) | Fault-to-fix histories for specific aircraft over many years |
| Medical imaging | Images paired with the radiologist's signed report (see medical imaging data) | Expert findings linked to the image they describe |
| Mortgage lending | Underwriting conditions, quality-control findings, pipeline history (see mortgage lenders) | Real decisions on real files, with outcomes |
| Call centres | Recordings and transcripts linked to the case outcome (see call recordings) | How problems were actually resolved, in real customer language |
| Construction | RFIs, submittals, change orders, daily logs, estimates against final cost (see construction companies) | Each project ends with a checkable cost and schedule result |
| Engineering | Native CAD with revision history, change orders, simulation and test results (see CAD and engineering data) | Design history lives only in the company's own PDM or PLM vault |
| Software | Source code with commit history linked to tickets (see source code) | Private repositories show how a real product was built and fixed |
What is operational data?
Operational data is the record a business creates by running day to day: orders, tickets, schedules, work orders, inspections, claims, invoices, messages and the systems that hold them. It is usually the largest share of a company's proprietary data, and the part AI developers ask about most, because it shows how work is actually done, step by step, with the outcome attached.
The phrase also has a narrower IT meaning. An operational data store (ODS) is a database that "is used for operational reporting and as a source of data for the enterprise data warehouse" (Wikipedia). An ODS holds some operational data; most operational data also lives in ticket systems, shared drives, email, CAD vaults and line-of-business software that no warehouse ever sees.
- For AI, the step-by-step record usually matters more than the summary. A dashboard of averages shows what happened; the underlying tickets show how.
- Origin's guide to licensing company data covers who helps prepare operational data for a licensing discussion and what each party decides.
| Kind of data | What it is | Example |
|---|---|---|
| Operational data | Records produced while doing the work | A work order with the technician's notes, parts used and time to fix |
| Analytical data | Data reshaped for reporting and analysis | Average time to fix by site and month |
| Master or reference data | The lists other records point to | The customer, asset or parts catalogue |
| Public data | Data anyone can access | Published standards, government statistics, a public website |
What does not count as your proprietary data?
Data you hold is not automatically yours to use or license. Three kinds need care before anyone calls them proprietary.
- Data licensed in from others, such as credit reports, market data feeds or purchased lists. Their license terms decide what you may do, and they rarely allow passing them on.
- Data that belongs to a client or platform customer. On collaboration tools the account holder usually owns the content: Slack's user terms, for example, say messages and files are Customer Data "owned by Customer", the organization that bought the service (Slack).
- Personal data inside your records. You may control a file, but people's personal data in it stays protected by privacy law. GDPR defines personal data as "any information relating to an identified or identifiable natural person" (GDPR Article 4). See de-identifying company data before licensing.
Why AI companies want proprietary data
AI developers want proprietary data because it holds what public data lacks: real decisions, the reasons for them and the results, recorded inside working businesses. Models trained or tested only on public text have seen what people publish, not how a claim was adjusted or a part was reworked.
Public deals show the market is real, though most licenses stay private. All figures are as reported or filed.
- These are large or unusual deals, not a price list. What a buyer offers for a company's records depends on the records, the rights, demand and the agreed scope; see what AI companies pay for company data.
| Licensor | Proprietary data | Reported terms |
|---|---|---|
| Reddit (S-1, February 2024) | User posts and data access | Data licensing arrangements with "an aggregate contract value of $203.0 million" and two to three year terms (TechCrunch) |
| Tempus AI (8-K, October 2026) | De-identified neurology records | License fees of "$100 per delivered record" (Stock Titan) |
| Spirit Airlines, in bankruptcy (2026) | Internal business records, including 100 million emails and 500 million workplace messages | Google won an auction for $10 million, subject to court approval and privacy objections (KLTV) |
How to protect proprietary data and still license it
You can license proprietary data and keep it protected by sharing a prepared copy under a written license that limits use, bars onward disclosure and requires deletion at the end. Licensing grants permission to use; it does not transfer ownership unless the contract says so. See what data licensing is.
- Keep the reasonable measures in place: access controls, confidentiality terms with staff and advisers, and markings on what is proprietary.
- Describe before you disclose. A first discussion can work from categories, years and volumes; no files need to leave the company.
- Remove what you cannot license: client-owned material, third-party data and personal data that has not been de-identified.
- Set the terms in writing: permitted use, exclusivity, security, no re-identification, and what happens to models after the term. The data licensing agreement checklist covers each clause.
- Origin Data Partners is an independent data partner. It helps owners describe their records and consider a fit with AI data buyers; a buyer may pay Origin when a deal closes. See what your data is worth from company-level answers. An estimate is not an offer.
Common questions
What is proprietary data?
Proprietary data is information a company created or collected, controls and keeps from the public, such as operating records, customer histories, designs, source code and pricing. It is valuable because others cannot get it on the same terms. Some of it also qualifies as a trade secret.
What is an example of proprietary data?
A manufacturer's work orders and inspection logs, an insurer's claim files, an airline's maintenance records, an imaging centre's images with signed reports, a lender's underwriting decisions, a software company's private source code and a call centre's recordings with case outcomes are all proprietary data, as long as the company created them and keeps them private.
What is proprietary information?
Proprietary information is company information that is owned and protected from disclosure: financial information, data, trade secrets, research and development, product designs, marketing plans, client lists, computer programs, processes and know-how. The US government glossary definition adds that it must be developed by the company, marked as proprietary and not available without restriction from another source.
Is proprietary data the same as a trade secret?
Not always. All trade secrets are proprietary, but proprietary data is a trade secret under US law only if the owner takes reasonable measures to keep it secret and it gets independent economic value from not being generally known (18 U.S.C. § 1839).
Is proprietary data confidential?
Usually, but the words mean different things. Proprietary means the company owns or controls the data; confidential means someone has a duty not to disclose it, usually under a contract or employment duty. Proprietary data shared without confidentiality terms can lose its protection.
What is operational data?
Operational data is the record a business produces by running: orders, tickets, schedules, work orders, inspections, claims, invoices and messages. It is the largest share of most companies' proprietary data and the part AI developers ask about most, because it shows how work is done and what resulted.
What is the difference between operational data and an operational data store?
Operational data is the business record itself. An operational data store (ODS) is a database that integrates some operational data for reporting and feeds the data warehouse. Much operational data never reaches an ODS: it sits in ticket systems, email, shared drives and specialist software.
Is proprietary data CUI?
Only when the US government holds it. Company information provided to a federal agency is controlled unclassified information in the category General Proprietary Business Information, marked PROPIN. The same data inside your own company is not CUI because of that label, though contracts with the government can add handling duties.
Can a company sell or license its proprietary data?
Yes, if it holds the rights and can remove what it may not share, such as client-owned material, third-party data and personal data. Most deals are licenses rather than sales: the buyer gets a defined right to use a prepared copy, and the company keeps ownership.
How much is proprietary data worth?
There is no standard price. Public examples range from $100 per de-identified medical record to multi-million-dollar archive deals, but value depends on years of history, how records link decisions to outcomes, clean rights and buyer demand. A company-level estimate is a starting point, not an offer.
Sources and further reading
Origin Data Partners is independent and not affiliated with any buyer named on this page.
- NIST CSRC glossary: proprietary information (from CNSSI 4009-2022)
- National Archives CUI Registry: General Proprietary Business Information
- 18 U.S.C. § 1839 (Defend Trade Secrets Act definitions), LII
- Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340 (1991), Justia
- Wikipedia: Operational data store
- GDPR Article 4: definitions
- Slack: User Terms of Service
- TechCrunch: Reddit says it has made $203M so far licensing its data (Feb 2024)
- Stock Titan: text of the Tempus AI 8-K (Oct 2026)
- InvestigateTV via KLTV: privacy protections sought in Spirit data sale to Google (Nov 3, 2026)