Can a software or SaaS company license its data for AI training?
Yes, if your company owns the material and has the rights to license it. A software company writes down how it ships a release, handles an outage, answers a billing dispute and why each service was built the way it was. Much of a company's data already lives in third-party software. Licensing it means re-licensing what already exists, and getting paid for it rather than giving it away for free. For you, that software is your wiki, your help desk, your issue tracker and your Git host. No new data collection is needed.
The gate that matters most here is the line between your own records and the content customers store in your product: the first can usually be licensed, the second generally only with their agreement.
Which software and SaaS data is worth licensing?
The material that shows how your team actually works is worth the most:
- SOPs and runbooks. On-call runbooks, incident playbooks, release checklists and support escalation procedures. See licensing SOPs and process documentation.
- Internal documents and knowledge bases. Product requirement documents, design docs, RFCs and post-mortems. See licensing internal documents and knowledge bases.
- Technical documentation. API references, architecture overviews and integration guides. See licensing technical documentation and specs.
- Support and customer-success conversations. Tickets, chat transcripts and escalations, with the resolution attached. See licensing support tickets and customer conversations.
- Code and engineering history. Repositories with their commits, pull requests, review threads and linked issues. See licensing codebases and engineering history.
- Engineering chat and email. The Slack or Teams channels where incidents were debugged and decisions argued out. See licensing chat and email archives.
Where does this data live in a software company?
It already sits in the tools your teams use every day, so nothing new has to be collected: documents and specs in Confluence, Notion or Google Drive, tickets in Zendesk, Intercom or a similar help desk, code and reviews in GitHub, GitLab or Bitbucket, and issues in Jira or Linear. What you decide is the scope: what goes in and what stays out.
Why do AI labs and agent builders want software company data?
Because agents are trained and tested on real tasks that follow real rules, and software companies hold both. SWE-bench, a coding-agent benchmark first published in October 2023, is built from 2,294 problems "drawn from real GitHub issues and corresponding pull requests" across 12 Python repositories (SWE-bench, arXiv). τ-bench, published in June 2024, tests agents in conversations with simulated users, each agent "provided with domain-specific API tools and policy guidelines", and found that even the state-of-the-art function-calling agents of the time, such as GPT-4o, succeeded on under half of its tasks (τ-bench, arXiv). A support SOP is exactly such a policy guideline, and a resolved ticket shows it applied.
Buying data from shut-down companies took off once AI training labs ran short of public and synthetic data, and one AI data company buys such data to build "reinforcement-learning gyms" where agents practise tasks such as debugging a complex IT architecture (MIT Technology Review, Oct 13 2026). Your private runbooks, tickets and review history record tasks of that shape, and only your company holds them. The licensees are AI labs and teams building AI models and agents.
What do your customer contracts and DPAs allow?
They usually decide whether customer content can be licensed at all, and the general answer is no, not without the customer's agreement. This is general information, not legal advice. Under the GDPR, a contract with a processor must stipulate that the processor "processes the personal data only on documented instructions from the controller" (GDPR, Art. 28(3)(a)), and a processor that infringes the Regulation by determining the purposes and means of processing is considered a controller in respect of that processing (Art. 28(10)). If your product stores your customers' documents, tickets or records under a DPA, you typically hold them as their processor.
Your own SOPs, specs, code and support conversations are a different category. Even there, check customer confidentiality clauses and what your privacy notice told people their data would be used for. The guide to whether selling company data to AI is legal covers the wider rules by region.
Does open-source code in your repositories change what you can license?
Yes: code you did not write stays under its authors' licences, so it is normally left out of the scope. The GPL is the clearest case. Version 3 requires anyone who conveys verbatim copies of its source code to "give all recipients a copy of this License along with the Program", and states "You may not impose any further restrictions on the exercise of the rights granted or affirmed under this License" (GNU GPL v3, s. 4 and s. 10). So the usual approach is to exclude vendored dependencies and third-party files, and license the code your company wrote.
Who owns the code and documents your employees wrote?
Employee work generally belongs to the employer; contractor work depends on the contract. In the US, a work prepared by an employee within the scope of his or her employment is a work made for hire, and the employer is the initial owner unless it has signed a written agreement to the contrary with the creator; a commissioned work counts only if it falls in one of nine listed categories and a written agreement signed by all parties says so (US Copyright Office, Circular 30). In the UK, the employer is the first owner of copyright in a work made by an employee in the course of employment, subject to any agreement to the contrary (Copyright, Designs and Patents Act 1988, s. 11). For agency and freelance work, look for a written assignment.
What is software and SaaS company data worth?
Partnerships typically range from $50K to $3M+. Public reference points are scarce and none is a fearofgod.ai price. One founder winding down a startup netted roughly $25,000 for its codebase, plus a pending amount likely around $15,000 for its workspace data in Google, Notion, Figma and Slack (MIT Technology Review, Oct 13 2026). Valuations vary: one firm that resells startups' data values it by head count, years in operation, volume of data and how the team used its tools (MIT Technology Review, Oct 13 2026). The guide to selling data to AI companies covers more reference points.
How does licensing work with fearofgod.ai?
You license what your company already keeps, and you approve the scope, so a customer's content, a sensitive service or a period of history can stay out. fearofgod.ai de-identifies the data before it goes to buyers. Raw data is deleted according to a schedule agreed with the partner. Your company can keep ownership of its data. This is agreed case by case. Your company's name stays confidential until the agreement is fully signed, and a completely anonymous partnership is also possible.
The steps are the partnership form, a conversation, scope and terms, data preparation, then delivery and payment. fearofgod.ai works with 60+ partner companies based in the US, the EU, the UK, Canada, Australia, New Zealand and Japan, and $17M has been paid to partners so far. The guide to data licensing agreements for AI training explains the terms you will see. Talk to our team when you are ready.
Selling software and SaaS company data to AI companies
A software or SaaS company that owns its SOPs, documents, support conversations and code can sell them to AI companies by licensing them with fearofgod.ai, securely and for long-term revenue. Partnerships typically range from $50K to $3M+. Partners can be paid more than once, through renewals and repeat licences.