Can You License Your Company's Codebase or GitHub History to AI Labs?

What in a repository can be useful

A snapshot of source code shows what was built. The history shows how: commits, pull requests, review comments, linked issues and test changes record real engineering decisions. That sequence is useful for training and evaluating coding assistants and agents, especially in less common languages or domains.

• Commits with meaningful messages
• Pull requests with review discussion
• Issues linked to the fixes that closed them
• Tests and CI results that show what worked

Ownership, contributors and trade secrets

Employee work is usually owned by the employer, but contractor code depends on the contract. Open-source dependencies keep their own licences and should be excluded or handled under those terms. Customer-specific code may be restricted by client agreements.

Licensing code exposes how your product works. Decide what counts as a trade secret, what can be excluded, and whether secrets, credentials and personal data in commit history must be scrubbed before any sample is shared.

Why there is no public price for code

Unlike images or text, there is no widely reported per-line or per-commit rate. Published dataset ranges from programmes such as Scale AI and micro1 describe general data partnerships, not code specifically, so they are not a guide to code prices. Value depends on language, domain, history depth, review quality and rights clarity.

How to prepare without giving anything away

Start with a written description: languages, age, number of contributors, how history is structured and what is excluded. Only share samples under a written agreement. Never upload a full repository for a speculative estimate.