What to put in an AI project contract: the six-clause baseline
An AI project contract does not need to be long, but it does need to cover six things clearly: scope and deliverables, payment, who owns the code and the model outputs, data handling and confidentiality, how revisions and changes are handled, and how either side can end the engagement. Get those six right and you prevent the large majority of client disputes before they can start. Most contract pain does not come from a missing clause about something exotic. It comes from a fuzzy clause about something basic.
I have spent fifteen years shipping inside teams at Apple, PlayStation and Schwab, and I learned that the contract is not there for the good projects. It is there for the day something goes wrong, and on that day a clear document is worth more than a good relationship.
The Six-Clause Baseline: the clauses every AI build needs
The Six-Clause Baseline is our CCC starting point for an AI project contract. It is not an exhaustive legal template. It is the six clauses that, in practice, cause the most trouble when they are vague or missing. Cover these clearly and you have removed most of the risk.
- Scope and deliverables. Spell out exactly what you are building and, just as importantly, what you are not. A named list of deliverables and a short list of exclusions is what you point to when a request drifts.
- Payment schedule and milestones. State the total, the schedule, and what triggers each payment. Tie payments to milestones so you are never far ahead on work and behind on cash. Include your terms for late payment.
- Who owns the code and the model outputs. Say plainly who owns what, and when ownership transfers. AI builds add a wrinkle: outputs, prompts, and any custom tooling all need an owner named. Often ownership transfers on final payment.
- Data handling and confidentiality. Set out whose data is involved, how it is stored and used, what stays confidential, and what happens to the data when the project ends. This clause protects both sides and is easy to forget.
- Revisions and change requests. Define how many revisions are included, and state that new requests are handled as priced changes. This is your Change Line written into the contract so scope creep has a home.
- How either side can end it. Describe how the client or you can end the engagement, what notice is required, and how work done up to that point is paid for. A clean exit clause turns a bad breakup into an orderly one.
The two clauses people get wrong on AI builds
Two of the six clauses cause more grief on AI projects than the rest combined: ownership and data. On ownership, the trouble is that AI builds have more moving parts than a normal project. There is the code, but there are also the prompts, the configuration, any custom tooling around the Claude Code workflow, and the outputs the system produces. Each needs an owner named. Silence here is how two parties end up both believing they own the same thing.
On data, the trouble is that AI work almost always touches the client's information, and clients care deeply about where it goes. Be explicit about how their data is handled, that it stays confidential, and that it is not retained or reused beyond the project. Getting this clause right is not just legal hygiene. It is a trust signal that wins you the work in the first place.
Tying payment to milestones so you never carry the risk
The payment clause is where you protect your cash. Tie each payment to a milestone the client can see and sign off on, so you are always paid for a stage before you begin the next one. A deposit before you start is standard and reasonable. It signals commitment on both sides.
Structure the schedule so you are never far ahead on delivery and far behind on payment. That gap is where agencies get burned, because leverage disappears the moment the work is done and the invoice is still open. Milestones keep you and the client moving in step.
Use directional terms if you must, but be concrete about the structure. A deposit, staged payments against milestones, and a final payment on delivery is a pattern that protects both sides and is easy for a client to accept.
Make the baseline yours, then compare notes
The Six-Clause Baseline is a starting point, not a finish line. Your specific work, clients, and jurisdiction will add clauses. But if you never ship an AI project without these six covered clearly, you have removed most of the ways a project turns into a dispute. Once more, and it matters: this is general guidance, not legal advice, so have a qualified professional review your contract.
Contract language is one of those things every agency owner refines over years of small painful lessons. If you are drafting or tightening your own AI project contract, bring the tricky clauses to the CCC community and compare notes on how others have handled ownership, data, and exit without scaring the client off.
Frequently asked questions
What is the Six-Clause Baseline?
The Six-Clause Baseline is a CCC starting point for an AI project contract. It covers scope and deliverables, payment and milestones, ownership of code and model outputs, data handling and confidentiality, revisions and change requests, and how either side can end the engagement.
Do I really need a contract for a small AI project?
Yes. The contract is not there for the projects that go well, it is there for the day something goes wrong. Even a short document covering the six baseline clauses prevents most disputes and costs far less than one bad breakup with a client.
Who should own the code and model outputs?
Whoever you and the client agree on, stated plainly in the contract. On AI builds you must name an owner for the code, the prompts, any custom tooling, and the outputs, since each is a separate thing. Ownership often transfers to the client on final payment.
What should the data clause cover?
It should state whose data is involved, how it is stored and used, what stays confidential, and what happens to the data when the project ends. Being explicit that client data is not retained or reused beyond the project protects both sides and builds trust.
How should I structure payment in an AI project contract?
Tie payments to milestones the client can sign off on, usually with a deposit before you start and a final payment on delivery. Structure it so you are never far ahead on work and behind on cash, because that gap is where agencies get burned.
Is this contract guidance legal advice?
No. This is general guidance based on what tends to go wrong on AI builds, not legal advice. Every situation and jurisdiction is different, so you should have a qualified legal professional review any contract before you rely on it or sign it.
Last reviewed by Duncan Rogoff on July 4, 2026
