Dev.to AI 🤖 Ai 👁 0 📖 2 min read

Your old private repos can earn money from AI training (and you keep ownership)

Most developers I talk to (I'm a tech recruiter, so that's a lot of developers) have the same thing on their disk: a handful of private repos that will never ship. An abandoned startup. An internal tool from a job that e

Most developers I talk to (I'm a tech recruiter, so that's a lot of developers) have the same thing on their disk: a handful of private repos that will never ship. An abandoned startup. An internal tool from a job that ended. A side project that hit 80% and stalled.

That code still has value, just not where you'd expect it. AI labs need large amounts of real-world, non-public code to train and evaluate coding models, and public GitHub has been scraped to death. So there is now a market for licensing private code, and you don't have to sell or give up your repo to take part.

Here's what I've learned about how it works, what qualifies, and what to watch out for.

How code licensing for AI training works

The model is simple:

  1. You connect your Git account and get a per-repo estimate.
  2. You choose which repos to offer. Nothing is shared without your approval.
  3. Secrets and personal data are stripped from the review copy before a buyer sees anything.
  4. A buyer licenses the code, and you get paid (typically within ~15 days of the buyer paying).

The important part: the license is non-exclusive. You keep ownership, you can keep using the code, and the same repo can be licensed more than once.

What it pays (indicative only)

Pricing is usually by volume of tokens:

What Indicative rate
Source code ~$100 per million tokens
Git history ~$40 per million tokens
Illustrative range ~$1,000–$1,500 per repo (not a quote)

These figures come from the program page and are not guaranteed. Rates can change, and deals take 2–16 weeks depending on buyer demand. Treat it as a bonus for code that's otherwise doing nothing, not as income you can plan around.

What qualifies and what doesn't

✅ Usually qualifies

  • Code you own or have the right to license
  • Private repos, internal tools you own, abandoned startups, side projects

❌ Doesn't qualify

  • Anything that was ever public, and forks
  • Client-owned or employer-owned code (check your contracts!)
  • Code containing confidential data or conflicting third-party rights
  • Generated files, vendored dependencies, binaries and data dumps (they don't count toward value)

Before you submit: a short checklist

  • Check ownership. If you wrote it at a job or for a client, it's very likely not yours to license. You give an ownership warranty, so take it seriously.
  • Check co-founders. For a dead startup, make sure whoever owns the IP (often the company) agrees.
  • Scan for secrets anyway, even though the platform strips them. gitleaks or trufflehog take a minute.
  • Read the full terms, especially on how the code may be used and on payout conditions.

Where to start

I put together a free guide with FAQs on GitHub: license-your-private-code.

If you want to try it, the program I refer developers to is Pangea:

Disclosure: those are my referral links. I may earn a commission if you license code through them, at no extra cost to you. The guide on GitHub is free and works without them.

Have you licensed code for AI training, or decided against it? I'd like to hear why in the comments, especially from anyone with concerns about how the code gets used.

📰 Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.