Yes. Legacy code is often the most valuable code a company holds, at least to an AI lab. A system that ran for eight or ten years carries a long record of how software really changes: features added under pressure, bugs fixed in production, frameworks upgraded, workarounds that became permanent. That record is hard to find and impossible to fake.
This page is for companies and teams holding systems that were retired, replaced or frozen. Think of the booking engine replaced in a platform migration, the in-house ERP nobody maintains any more, the monolith that was split into services, or the old version of a product that was rewritten from scratch.
The stack does not need to be current. PHP, Perl, Java, .NET, Rails, Delphi and COBOL all count. Older languages can even be scarce in training data, which makes them useful. What matters is that the code is real, that the history is intact, and that your company holds the rights.
How do I monetize old code nobody maintains?
Start with an inventory. Old systems hide in odd places: an archived organization, a self-hosted GitLab, a Subversion server under someone’s desk. Bring the repos together on a Git host you control, keeping the history when you migrate. Check the rights for each system, especially anything built with a vendor or a client. Then license the repos as they are. Nobody needs to maintain, document or deploy them first, and nobody else needs to know they exist.
Retired systems often carry the most history a company owns. A decade of commits from dozens of developers is far more than a new product will have for years. That history is what labs pay most for, at $40 per million tokens.