Whether it’s a quick question or a detailed brief — we’d love to hear about it.
AI Hub 032 · Guide
Who owns the code, the data and the IP when an outside team builds your AI?
If an outside team builds your AI, who owns it when you part ways? You should: the code, the rights in it, your data, the accounts it runs on and the know-how to keep it working. But paying for the work does not make it yours by default. In both the UK and the US, the firm that writes the code owns it unless a written agreement says otherwise.
So ownership is two things: what the contract says, and what you can check in the first week. It is the question we expect before anyone starts a dedicated AI development team with us. Sources checked on 17 November 2026. This is general information, not legal advice, and the rules differ by country, so have a lawyer read your contract.
Why paying for it does not make it yours
In the UK, the Intellectual Property Office says that when you commission another person or organisation to create a work, the first legal owner of the copyright is whoever created it, not you, unless you agree otherwise in writing. A transfer of ownership only counts if it is in writing and signed by the party giving up the rights.
In the US, a commissioned work is only a “work made for hire”, owned by the party that ordered it, if it falls into one of nine listed categories and both sides sign a written agreement saying so. The US Copyright Office’s list covers things such as translations, compilations and parts of a film. Computer programs are not on it. So do not rely on the words “work made for hire” alone. Ask for a written transfer of the rights as well.
Five things to own, and how to check each
1. The code
The contract should transfer all rights in the work to you, in writing. The check: the repository sits in your organisation’s account from the first day, not handed over at the end. If you can see every change as it is made, you already hold the code.
2. Everyone’s rights in it
A transfer only covers the people who signed it. Ask whether freelancers or subcontractors will touch your work, and whether they are bound by the same terms. Ask what the builder brings that it already owned, such as its own tools, and what right you have to go on using them.
3. Your data
You decide what data an AI tool may see and where it is processed. If the work involves personal data, UK data protection law keeps you responsible for it. The ICO says you need a written contract with anyone who processes it for you, and that the contract must have them act only on your documented instructions, keep the data confidential and secure, and delete or return it at the end.
4. The accounts and keys
Cloud hosting, AI model accounts, domains and monitoring should be in your name, or handed over with the code. Each person on the team should have their own access, so it can be removed one person at a time.
5. The know-how
Code you cannot run is not much of an asset. Ask for the documentation, the prompts and rules the tool works to, the tests that show it is working, and the steps to release a change.
Two questions AI adds
Does the AI company train on your data?
It depends on the service and the plan. OpenAI says that by default it does not train its models on data sent through its API platform. Anthropic’s commercial terms say it may not train models on customer content. Google’s Gemini API terms draw the line by plan: on paid services it does not use prompts and responses to improve its products, and on unpaid services it does, and human reviewers may read them. So ask which accounts your builder will use for your work, and make sure none of them is a free tier.
Who owns code that AI helped write?
In the US, the Copyright Office set out its position in January 2025: copyright protects what a human author contributed. Material generated wholly by AI is not protected, and prompts alone do not make someone the author, but using AI as a tool does not take protection away from the human’s own work. Other countries may differ. The practical question is the same everywhere: who reviews, edits and answers for what the AI writes?
What to get in writing before work starts
- A signed transfer of all rights in the work to you, covering everyone who works on it.
- What the builder keeps, and your right to go on using it.
- The repository and the accounts in your name from the first day.
- A data processing agreement, if personal data is involved.
- Which AI services will be used, and on which plans.
- What happens at the end: the code, documentation and accounts handed over, your data deleted or returned, and their access removed.
How we handle it
Everything the team builds belongs to you from day one. You decide what data the AI can access and where it’s processed. Every change is peer-reviewed and security-scanned before it reaches you.
You have the repository and the project board from the start, and you can join the daily standup. Tools and agents produce the volume, and one named developer answers for what ships. See how we work.
Requests to AI models go through a gateway we built. Each developer and agent has their own key, which can be revoked without affecting anyone else, and your code stays on the developer’s machine, because the gateway only relays model requests.
Whichever way you buy, you end with the code, the documentation and the accounts. We share the playbooks with our clients too.
Ask this one before you sign
Ownership is one of twelve questions worth asking any partner. The full list is in choosing an AI development partner. See how a dedicated AI development team works with us, or book the free consultation and ask us this one first.
Offshorly is a UK-led AI development company that builds dedicated AI development teams in the Philippines, trading since 2014.