Can a software company license its code to AI data buyers?
· Company Data Advisors
12 of the 17 AI data-buyer programs we track name software or code. They want private code with its history. The code is often yours to license; customer data often isn't.
Usually yes. As of November 12, 2026, at least 12 of the 17 programs in our buyer index name software companies, code or engineering records on their own pages, and Labelbox Prism marks software and internet as "urgent." What they want is narrower than "a codebase": private code that people wrote over several years, with the pull requests, reviews and tickets that show how it changed. In our read, for most software companies the code and the engineering history are the easier part to license. The harder part is everything that holds customers' information, above all the production database of a B2B product.
This is a reading of buyers' public pages and public legal sources, not legal advice. Have counsel review your customer contracts and code ownership before you share anything with a buyer, including a sample.
What do AI data buyers want from software companies?
AI data buyers want the record of software being built: source code with its git history, pull requests and code reviews, the tickets behind each change, CI logs and the design docs and postmortems around them. Mercor has a Software Company listing that asks for "private repositories that many engineers have worked on for years, with tests that still run." Collinear asks for "private repositories with pull requests and linked tickets" and says it is "especially interested in several years of work from multiple contributors." Labelbox Prism lists software and internet first among its industries, marked urgent, covering "engineering, product, sales, and ops workflows across tools like GitHub, Jira, Slack, Salesforce, and Snowflake."
| Buyer | What it names for software companies | Size or history it states |
|---|---|---|
| Mercor | Private code with its history (Software Company listing); product databases, warehouse and analytics code (SaaS listing); also AI product usage data and security records | 15 to 500 employees and 3+ years of commits for code; 50 to 1,000 employees with a data team for the SaaS listing |
| Collinear | Private repositories with pull requests, linked tickets, CI logs and coding-agent sessions, plus business app history | About 50 people or more, several years of history; smaller teams with deep records can work |
| Turing (Project Lazarus) | Source code, reviews, tickets and the iteration history behind what shipped; software and engineering listed first among its strengths | Not stated |
| Labelbox Prism | Software and internet marked urgent; code generation and edit traces (IDE sessions, diffs, test results, review comments) marked high demand | Not stated |
| micro1 | A software and engineering category: code documentation, engineering workflows, bug tracking, product planning and technical knowledge bases | 30+ full-time employees |
| Centific | Service and incident records, "particularly software & technology companies and managed IT environments" | Not stated |
| Protege | Engineering issues and reviews; says it moves fastest on ServiceNow IT ticket-resolution data | Not stated |
| Scale | Software and internet among its industries; Jira, Confluence and Linear tickets and wikis with their full history | Not stated |
| SourceX | Software among its industries and software engineering records among its data types | Not stated |
| Replay | Code and commits among the records it lists | Not stated |
| Handshake | Software, with technical documentation, issue histories and development workflows | 20 or more full-time employees and 3 or more years in operation |
| Polyshares | Software listed first among its sectors | No minimum stated |
Sources: each buyer's own pages, read on 2026-11-11 and 2026-11-12. Each buyer's page in our index links to its program. The table covers what buyers say about software; most of them also take the Slack, email and CRM records any company holds.
Two differences between programs matter. Turing's FAQ says source code is not required ("Communication, tickets and docs often matter more") but that you can offer code on its own, and that "Code earns most with its history: commits, pull requests and reviews, tickets and docs." Mercor's SaaS listing is the most explicit about databases: it asks for "the company's own production databases and a cloud warehouse." Collinear lists database dumps among the records it licenses, and Turing's FAQ includes databases. That raises the customer-data question covered below.
What makes a codebase worth more to buyers?
Buyers pay more for code that shows people working on a real product over time. Their published criteria line up closely. Mercor's Software Company listing asks for:
- private repositories "that have never been public, mirrored or open-sourced";
- 3+ years of commit history from at least 10 engineers;
- 50,000+ lines of code "the company owns outright";
- "Code written by people, with tests that run."
Mercor lists what "adds to what the company earns":
- thousands of pull requests with review discussion;
- tickets, design docs, runbooks and postmortems from the same team;
- code in C, C++, Rust, Swift, Kotlin, Ruby, COBOL or .NET;
- banking, insurance, healthcare, trading or blockchain software;
- history that "mostly predates 2025";
- openness to an exclusive license on part of the code.
Collinear's list of "what raises the price" is close: fifteen or more people at peak, several years of history, pull requests with the tickets behind them, human-written code, and "the same work visible across several tools."
The two programs also say what they don't want. Mercor rules out forks or copies of open-source projects, code mostly written by AI tools, client work an agency doesn't own, and single-author repositories. Collinear excludes "AI-generated code, or a public framework with a thin layer on top" and "anything you do not hold the rights to."
Mercor publishes a range for the Software Company listing: the company earns $50k to $500k, "more for large codebases or exclusive terms."
Our view: the preference for human-written history before 2025 follows from how buyers use code. Collinear says it builds "the tasks and simulated environments that AI agents train and are evaluated in." A coding agent learns from the path from ticket to merged fix, including the review comments and failed tests along the way, and that path is only useful if a person walked it. Our post on flight simulators for office work explains how records become those practice environments.
Do you own all the code in your repositories?
Probably not all of it, and buyers ask. Mercor wants code "the company owns outright," and Collinear excludes anything you don't hold the rights to. Before you describe a codebase to any buyer, sort the code by where it came from.
- Code written by employees. Under the Copyright Act, code an employee writes within the scope of employment is a work made for hire, and the employer is the author and owner, according to the U.S. Copyright Office's Circular 30.
- Code written by contractors and agencies. Circular 30 says commissioned work is a work made for hire only if it falls in one of nine listed categories and a written agreement signed by both parties expressly says so. Computer programs are not one of the nine. Whether you own contractor code depends on what your contracts say, which is a question for counsel.
- Code you built for clients. If you are an agency or consultancy, client contracts may give the client the code. Mercor names "client work an agency doesn't own" as not a fit, and its SaaS listing rules out "agencies holding their clients' databases."
- Open-source code. Buyers exclude forks and thin layers over public frameworks. Open-source libraries copied into your repositories aren't yours, and each one's license decides what you can pass on. Know roughly how much of each repository they make up.
- Code from acquisitions. Mercor accepts "codebases from acquisitions," as long as the code was never public. Check that the purchase agreement transferred the code and doesn't restrict its use.
At the buyers that address it, a license leaves ownership with you. Collinear says "Ownership stays with you," and Turing says "Ownership doesn't change hands." Turing also says a future acquirer's diligence "may ask about it, so loop in your counsel before you sign," which matters for a software company that may be sold.
Can a SaaS company license its customer data?
In our view, usually not without its customers' agreement, because a B2B software company usually holds its customers' data on their behalf. Under the EU's GDPR, such a company is usually the processor and its customer the controller, and Article 28 requires the contract between them to say the processor "processes the personal data only on documented instructions from the controller." Under Article 28(10), a processor that breaches the Regulation by determining the purposes and means of processing "shall be considered to be a controller in respect of that processing." California's privacy law, the CCPA, defines a service provider (Civil Code 1798.140(ag)) as one whose contract prohibits it from selling or sharing the personal information, or retaining, using or disclosing it "for any purpose other than for the business purposes specified in the contract."
Those are the privacy rules. Our view: your customer contracts usually add confidentiality terms that reach beyond personal data to everything the customer puts into your product. Anonymizing the records may satisfy a privacy rule and still breach a contract that says the data is the customer's and may be used only to provide the service.
In our reading, this bears directly on the records buyers name. A production database, a data warehouse, support tickets and AI product usage logs all contain what your customers and their users did in your product. Mercor's SaaS listing asks for "the company's own production databases and a cloud warehouse." For a consumer subscription business that data may be the company's own, subject to its privacy policy. For a B2B product, much of it is the customers'.
Our view: sort a software company's records into the groups below and treat each differently.
| Records | Our read | Why |
|---|---|---|
| Source code, git history, pull requests, code reviews, CI logs | Candidate, once ownership is checked | Company work product; check contractor, client and open-source code first |
| Internal tickets, design docs, runbooks, postmortems, engineering chat | Candidate, after review | Company records, but they can quote customer data, name customers or include incident details |
| Secrets in repositories (keys, passwords, connection strings) | Remove before any sample | Turing says "Personal data and secrets are removed before anything reaches a model," but rotate anything exposed in history anyway |
| Support tickets and customer conversations | Check contracts first | Written by or about customers; may be covered by confidentiality or data processing terms |
| Production databases, warehouse tables, product analytics, AI usage logs | Assume it's the customers' unless contracts say otherwise | Usually processed for customers under contract; see Article 28 and the CCPA service-provider rule above |
| Employee details in commits and chat | Expect pseudonymization; ask how | Commit authors and reviewers are named; Collinear says names are replaced with consistent pseudonyms |
This is a screening view, not a conclusion about any company. Game studios face extra limits from engine, console and middleware licenses; our guide to game studios covers them. Our guides to records that include employee and customer data and to Slack, Google and Microsoft terms cover the personal data and vendor terms in more depth.
What does licensing code mean for your competitive position?
A license gives the buyer the right to use your code for the purposes the agreement names, and the terms decide how far it travels. Three terms matter more for code than for most records:
- Exclusivity. Mercor counts openness to an exclusive license on part of the code as adding to what the company earns. In our view, an exclusive license usually limits who else you can license that code to, and depending on its terms it may also limit your own use. Read what the exclusivity covers, for which uses and for how long.
- Onward licensing. Turing licenses non-exclusively, "so the same record can go to more than one lab." Ask every buyer who will receive your code and whether you can block named competitors. Our comparison of license terms across buyers covers resale and permitted recipients.
- What leaves your systems, and when. Turing says that after the NDA it asks you to "run our repo evaluation script on your machine," and that "Your code stays with you until we agree on a sample." Collinear suggests a sample such as "a zipped repository with its git history" and signs a mutual NDA before samples on request. Ask what the evaluation reads and what a sample needs to include.
Our view: for a live software company, the code itself is rarely the competitive secret; the customers, the roadmap and the data are. Still, list the repositories you would never license (a pricing engine, a security product's detection logic) before you talk to anyone, and keep them out of the sample.
Where we come in
A software company that decides to license often has more than one fit: one buyer for the code, another for the tickets and engineering chat, and a few asking for data it may not be free to share. We track the published criteria and terms of every major program, have direct contacts at the major data buyers and at AI labs, and read the ownership, exclusivity and onward-licensing clauses with you before you choose. Nothing about your company is shared until you approve a specific, named buyer. Participating buyers may pay us referral fees, and we tell you which before any introduction. Owners pay us nothing, and you can apply to any buyer directly.
What to do next
Before you talk to a buyer, put four lists in front of your CTO and counsel:
- Repositories: each private repository, its first commit date, the number of human contributors, and whether any of it was ever public.
- Ownership: which code came from contractors, agencies, clients, open-source projects or acquisitions, and the contract that covers each.
- Customer data: which systems hold customers' data, and what your customer agreements and data processing terms say about using it for anything other than the service.
- Never license: the repositories and records you would keep out of any deal.
Then compare programs on the records you can actually license. For the steps from first form to payment, see how to sell or license your company's data to AI companies, or check where you fit.
Programs in this post
More from the blog
- Is my company too small to license its data to AI buyers?
Most AI data buyers with a size floor set it at 20 to 50 employees, but law, accounting and medical firms qualify smaller, and nine programs state no minimum.
- What company data AI buyers want: November 2026
Across the 17 AI data-buyer programs we track, accounting, healthcare and software lead the industries, and CRM, ticket, email and finance records lead.
- One payment or a revenue share: how AI data buyers pay
The largest group of AI data buyers pays once. A few pay a share of every resale, and one now pays nothing upfront. How each model works and what to ask.
See which buyers fit your company
Eight quick questions against every program in our buyer index. Participating buyers may pay us referral fees.