Key takeaways
- Client AI clauses are becoming common in professional services, government and financial services contracts.
- Most can be met easily if you use business-tier tools with recorded data commitments and human review.
- Push back on blanket bans with a narrower, workable clause.
- Ask your insurer about AI before renewal, and put your own AI terms in your engagement letter.
Two documents are starting to ask about AI: your clients' contracts and your insurer's renewal forms. Neither is a reason to slow down. Both are a reason to have your setup in order, because the businesses that can answer the questions clearly win the work and the cover, and the ones that cannot end up agreeing to terms they cannot keep.
The clauses you will see
| Clause type | What it says | Usually workable? |
|---|---|---|
| Disclosure | Tell us if you use AI in delivering services | Yes |
| Approved tools | Only use AI tools with specified security and data commitments | Yes, with business-tier tools |
| No training | Our data must not be used to train AI models | Yes, business tiers commit to this by default |
| Human review | A qualified person must review AI-assisted output | Yes, and you should do this anyway |
| Data location | Data must be processed in a specified country | Depends on provider and tier |
| Prior consent | Ask before using AI on our matters | Workable but slow |
| Blanket ban | No AI use at all | Often unworkable in practice |
Why blanket bans do not work
AI features are now built into email, word processing, meeting software and search. A clause that bans "any use of artificial intelligence" can be breached by a spell checker or a meeting transcript. Agreeing to it means either breaching it routinely or turning off features across your business for one client.
Offer a narrower alternative instead:
Suggested wording to propose: "The Supplier may use AI tools in delivering the Services only where (a) the tool is provided under a business or enterprise agreement under which Client Data is not used to train models, (b) Client Data is handled in accordance with the Supplier's AI usage policy, a copy of which is available on request, and (c) all deliverables are reviewed by an appropriately qualified person before being provided to the Client." Have your lawyer adapt this to your contracts.
What you need in place to say yes
- Business-tier AI tools with recorded data commitments. See choosing the right AI plan tier.
- No staff using personal accounts for client work. See moving off personal ChatGPT accounts.
- A written usage policy you can share. See AI usage policy for teams.
- A review step that is actually followed.
- A list of approved tools you can give to a client on request.
Insurance
Professional indemnity policies generally respond to negligent professional services, whatever tools were used. But insurers are starting to ask about AI at renewal, and some are considering specific terms. Before your next renewal:
- Ask your broker in writing whether AI-assisted work is covered on the same basis as other work.
- Ask whether any conditions apply, such as human review or approved tools.
- Answer AI questions on proposal forms accurately. Describe the tools, the controls and the review process.
- Check cyber cover too, particularly for agents with access to your systems.
Write your own AI terms first
The strongest position is to put a short AI statement in your own terms of engagement before clients ask. It sets the default on your terms, reassures clients, and shows you have thought about it. A good statement covers which tools you use and under what agreements, that client data is not used for training, that professionals review all work, and that clients can ask for more detail. We cover the client conversation in what to tell clients about AI use.
This article is general information, not legal or insurance advice. Have contract terms reviewed by your lawyer and confirm cover with your broker.
Frequently asked questions
What is an AI clause in a contract?
A term that sets rules on whether and how a supplier may use AI in delivering services. Common versions require disclosure, restrict which tools can process the client's data, prohibit training on the client's data, require human review, or ban AI use for certain work.
Should we agree to a client's ban on AI use?
Not automatically. A blanket ban can prevent you using ordinary features in your email and document software. Propose a narrower clause instead: approved business-tier tools only, no training on client data, human review of all output, and disclosure on request.
Does professional indemnity insurance cover AI mistakes?
It depends on the policy wording. Many policies cover negligent professional services regardless of the tools used, but some insurers are adding questions or terms about AI. Ask your broker directly whether AI-assisted work is covered and whether any conditions apply, and get the answer in writing.
Should we tell clients we use AI?
Yes, in a proportionate way. A short statement in your terms of engagement about how you use AI, what controls apply and that professionals review all work is usually better than either silence or detailed disclosure on every document.
Get a straight answer for your business
The Clarity Package gives you a written business case and a detailed implementation plan for 500 USD. The Implementation Package does the whole build, live in 30 days, for a fixed 5,000 USD.