A data title search is an examination, before a data deal closes, of whether the seller has the right to license the data. It traces how the data was created, who owns it, and what employee agreements, customer contracts, privacy notices and earlier licenses restrict. The result is a written title report.
Why data needs a title search
In a property sale, nobody takes the seller's word that the house is theirs to sell. A title search examines the deed and anything recorded against it. Data has no public register, but the same question applies. A company can hold a dataset for years without holding the right to license it to someone else.
The gap between holding and owning is where data deals fail. A support archive may be full of customer content that a contract says must stay confidential. A code base may include work by a contractor who never assigned the rights. A document store may contain licensed research that cannot be passed on. A data title search finds these limits before the buyer does.
The search builds a chain of title: a documented path from the creation of the data to the right to license it today. Each link is a question with evidence behind it. Outrich Data runs this search before any purchase it makes, and offers it to other buyers and sellers as part of its escrow service.
Creation and ownership
The first link is how the data came to exist. Records generated by the company's own operations, such as its internal tickets, logs and process documents, usually start with the cleanest position. Records supplied by customers, collected from users or bought from vendors start with someone else's rights attached.
The search sorts the dataset by origin. For each source system it asks who entered the records, under what terms, and whether the company is the owner or only a holder. A company that processes data on behalf of clients is often a service provider for that data, not its owner. That material normally comes out of scope unless the client agrees otherwise.
Corporate history matters too. If the data arrived through an acquisition, the purchase agreement shows whether the data and the related rights transferred. If the data sits in a subsidiary, the entity signing the license has to be the one that holds the rights.
Employee and contractor IP assignment
Much company data is written by people: documents, code, messages, analysis. In the United States, work an employee creates within the scope of the job generally belongs to the employer, but the detail varies by country and by the type of right. Most companies close the question with an invention assignment or intellectual property clause in the employment agreement.
Contractors are the weaker link. A contractor usually keeps the rights in what it creates unless a written agreement assigns them. The search looks for signed agreements covering the people and firms who contributed material, and notes any periods or teams where paperwork is missing.
Ownership of the work is one question. Employee privacy is another. Messages and recordings can carry personal information about the people who wrote them, and monitoring notices and local employment law may limit how that material is used. The search flags this as a separate item.
Customer contracts and data processing agreements
Customer agreements are the most common source of restrictions. The search reads the standard terms the company has used over time and any negotiated contracts with larger customers. It looks for confidentiality clauses, limits on use of customer data, clauses that say the customer owns what it submits, and audit or deletion duties.
Data processing agreements deserve their own review. A DPA usually says the company may process personal data only on the customer's instructions and for the purposes of the service. Licensing that data to a third party for model training is rarely within those instructions.
The outcome is a map. Some customers' records are clear. Some are blocked. Some could be cleared with consent or an amendment. A dataset can often still be licensed after the blocked accounts are removed, and the title report records which were removed and why.
Privacy notices over time
A privacy notice tells people what a company will do with their information. The notice that counts for a given record is generally the one in force when that record was collected, so the search reviews each version across the dataset's time span, not only the current one.
The review compares the stated purposes with the proposed license. It notes whether sharing with third parties was disclosed, whether any promise was made not to sell personal information, and which laws applied to the people involved. De-identification changes this analysis and can take data outside some privacy rules, but it does not cancel a contract promise. The notice review and the de-identification plan are read together.
Prior licenses and third-party material
The last links concern what has already been granted and what was never the company's to grant. An earlier exclusive license can block a new deal entirely. An earlier non-exclusive license does not block it, but the buyer will want to know about it because it affects value and any claim that the data is unseen.
Third-party material hides inside otherwise clean datasets. The search checks for the items below and recommends removal or a narrower scope where they appear.
- Earlier licenses, data-sharing agreements and any exclusivity or most favored terms in them
- Security interests or lender liens that cover company data as collateral
- Licensed content such as market research, standards, stock media and paid databases
- Open-source code and its license conditions inside a code repository
- Material received under a non-disclosure agreement
- Content scraped or copied from websites whose terms restrict reuse
- Records previously published, which affects held-out and contamination claims
What a title report contains
The title report is the written result. Both sides of a deal receive the same document. It does not say the data is risk-free. It says what was examined, what was found and what must change before closing.
Findings are usually sorted into three groups: clear to license, restricted, and to be cured before closing. A cure might be removing a customer's records, obtaining a missing assignment or narrowing the license scope.
| Section | What it states |
|---|---|
| Dataset description | Source systems, record types, time span and volume examined |
| Ownership basis | How the data was created and why the seller is the owner |
| People and IP | Assignment coverage for employees and contractors, with gaps noted |
| Contract review | Customer, vendor and partner terms that restrict use, by group |
| Privacy review | Notice versions, consent basis and laws considered |
| Prior grants | Earlier licenses, exclusivity, liens and publication history |
| Third-party material | Licensed, open-source or confidential content found |
| Exceptions and cures | What is excluded and what must be fixed before closing |
| Documents reviewed | An index of every document the findings rest on |
A title report is not title insurance
In property, a title search is usually followed by a title insurance policy that pays if a covered defect appears later. A data title report is not that. It is a documented examination, not an insurance policy, and it does not pay out if a claim arises.
The report is also limited to the documents provided. It cannot find a contract nobody disclosed. For that reason the seller still gives representations and warranties in the license, and the indemnity clause still allocates the risk of a later claim. What the report changes is how much of that risk is unknown. A buyer with a clear report has less to discount, and a seller who has read one knows exactly what it is promising. Whether a given dataset can be licensed remains a question for the parties' own counsel.