Chain of Title: Why Rounds Break on Code Ownership

Before a term sheet turns into money, the fund’s counsel will ask you to show how the rights to every part of the product moved from the people who created it to the company. That trail is the chain of title. Gaps in it delay rounds more often than weak metrics, because one missing signature can mean the company never acquired its main asset.

The question investors actually ask

“Do you own your IP” appears on every checklist, and every founder answers yes. The work starts with the follow-up: who wrote the first version of the engine, on what contract, and where is the document that moved those rights to the company? Picture the screen share during due diligence: the associate opens the repository history and points at hundreds of commits from a private email address, dated a year before the company was registered. If the author is a co-founder, the rights to that code still sit with them personally. Incorporation moved nothing. A line in the shareholders’ agreement saying “all IP belongs to the company” usually fails too, because it lacks what a copyright assignment needs.

Four places where the chain breaks

The same four gaps appear in most technology companies. First, founder code written before the company existed. Second, contractors. Under Polish law, an agreement transferring author’s economic rights must be made in writing or it is void (Article 53 of the Copyright Act of 4 February 1994). An invoice with an email thread transfers nothing. Employees are a separate case: Article 12 gives the employer the rights to works created within employment duties, and only within them, so the app someone built at a previous job stays outside. Third, open source. The codebase depends on hundreds of packages and nobody kept a license record; a copyleft component in the core product surfaces in every serious review. Fourth, assets from AI tools: the logo, marketing copy, fragments of code. Whether purely machine-generated output enjoys copyright protection at all remains uncertain, and the tool’s terms decide commercial use. If you cannot say which asset came from which tool, you cannot answer the fund’s question.

Cleaning up takes weeks, so start before the round

None of these gaps ends a deal if you find them first. Start with a map of creators: every person who committed code, designed assets or wrote content, with their legal status at the moment of creation (founder, employee, contractor, agency). Match each entry against a signed assignment. Where one is missing, sign a confirmatory assignment now: in writing, listing the fields of exploitation, signed by the actual author. This gets slow when a contractor has vanished or expects extra payment, so do it before the round, without term sheet pressure. Then scan your dependencies for licenses and keep the output as a living open source register. Finally, file everything in a data room folder before anyone asks: author map, contracts, annexes, license register, a log of AI-generated assets.

What we recommend now

Five steps fit into the next two weeks:

• Pull the full commit and asset history and list every author with their contract status at the time of creation. One spreadsheet is enough.
• Check each contractor file for a written assignment covering the fields of exploitation. Where it is missing, prepare confirmatory assignments and start collecting signatures, hardest-to-reach authors first.
• Run a license scan of your dependencies, save the result and decide what happens to any copyleft finding in the core product.
• Write a one-page rule for AI-generated assets: which tools are allowed, what gets logged, which materials need human rework or replacement.
• Open the IP folder of your future data room and file all of the above in it this month.

Sources: Polish Copyright Act (ISAP)