Insurance, liability and indemnity for independent AI consultants
6 minute read. Updated 2026-08-08.
Professional indemnity insurance is the core cover for independent AI consultants, but AI work raises the stakes because a flawed model can cause harm that compounds quietly across many customers, far beyond the fees you were paid for the work. Check that your policy explicitly covers AI-related work, read indemnity clauses separately from any liability cap, and take a contract you do not fully understand to a solicitor before you sign, not after.
What cover actually exists
Most independent AI consultants can name their professional indemnity policy but could not tell you what it excludes. That gap is worth closing before you ever need to make a claim. Professional indemnity insurance, PII, is the core policy for anyone building or advising on systems a client will rely on: it responds to claims that your work was negligent and caused financial loss, a wrong specification, a flawed model, a missed data quality issue that fed bad decisions downstream.
Public liability covers physical injury or property damage, rarely central to software work but sometimes demanded anyway because it sits in a standard contractor terms template. Cyber liability is increasingly relevant: it covers the costs of a data breach, including notification and remediation, which matters directly if you handle client data to build or evaluate a model. If you ever employ staff or subcontractors, employer's liability insurance becomes a legal requirement rather than a choice.
Why AI work raises the stakes
A conventional software defect usually has a bounded blast radius: a bug breaks a feature, someone fixes it, the loss is roughly the cost of the outage. A flawed model deployed into a decisioning process does not announce itself the same way. It can keep making the wrong call quietly, at scale, across every customer flowing through it, for months before anyone notices the pattern.
That makes the size of a potential claim far harder to bound at the point you sign the contract. You are pricing a day rate for the work you did, but the client's exposure, and by extension what they might claim from you, is priced on the harm the output caused, which has no fixed relationship to how many days you billed.
Insurers have noticed. Some professional indemnity policies now carry specific exclusions or sub-limits for AI-related work, or ask detailed questions about the models involved before quoting. Do not assume a policy written for general consultancy work covers AI-specific claims; check the wording, or ask the broker directly.
The unlimited liability clause
Client contracts commonly try to cap what you are liable for, expressed as a multiple of fees paid, a fixed sum, or, in the client's preferred version, left uncapped entirely. An unlimited liability clause means there is no ceiling on what you could owe if something goes wrong, regardless of what you were paid for the engagement.
Watch indemnity clauses too, and read them separately from the liability cap. Contracts sometimes carve specific obligations, breach of confidentiality, IP infringement, breach of data protection law, out of the general cap and make them uncapped indemnities instead. A liability cap that reads well in one clause can be quietly undone by an indemnity clause in another.
What a sensible position looks like
A liability cap tied to a multiple of fees paid, commonly a year's fees or the value of the specific piece of work, is the standard ask from an independent consultant, and most clients accept it once you raise it, because their own suppliers ask for the same thing.
Where a client insists on an uncapped position, that is a term to negotiate hard, not a formality to accept because they are bigger than you. It is reasonable to say plainly that unlimited liability is not something your business, or your insurer, can support.
Get advice, not opinions
This is mechanics, not guidance on what to sign. Take a contract with a liability or indemnity clause you do not fully understand to a solicitor who works with contractors, not a generalist, and take your insurance questions to a broker who covers AI or technology consultancy specifically, not a generic small business policy.
A read-through by a peer or a forum post is not the same thing as advice from someone who carries responsibility for getting it right. The cost of an hour with the right specialist is small next to the cost of finding out what an uncapped clause means after a claim has already been made.
Reviewing cover as the work changes
The cover that suited a data pipeline project may not suit a model evaluation or red-teaming engagement taken on afterwards. Review your policy whenever the nature of the work changes materially, not just at renewal.
Tell your insurer what you actually did, in plain terms, rather than what your contract template says you do. A mismatch between the description on file and the real scope of work is exactly the gap an insurer will use to decline a claim.
What to do about it
- Confirm your PII policy explicitly covers AI-related work; do not assume a general consultancy policy does.
- Read indemnity clauses separately from the liability cap; they can undo it.
- Push back on uncapped liability clauses; a cap tied to fees paid is a normal, defensible ask.
- Review your insurance whenever the nature of the work changes, not only at renewal.
- Take a liability clause you do not understand to a contract solicitor before you sign, not after.
- Do not rely on a peer's read of a contract as a substitute for advice from someone who carries responsibility for it.
Questions people also ask
Do I need professional indemnity insurance as a fractional AI consultant?
Most clients require it as a condition of the contract, and it is sensible even where they do not ask, because it responds to claims that your advice or work was negligent and caused financial loss. Confirm with your insurer or broker that the policy wording explicitly covers AI and model-related work, since some policies now exclude or sub-limit it. Do not assume a general consultancy policy extends to this without checking.
What does an unlimited liability clause actually mean in practice?
It means there is no ceiling on what you could be required to pay if a claim succeeds against you, regardless of the fees you were paid for the engagement. Because a flawed model can cause harm that compounds across many customers over time, this exposure is harder to bound in AI work than in most other consulting. Take advice from a solicitor before agreeing to it.
Can I negotiate a liability cap with a large client?
Usually yes. A cap expressed as a multiple of fees paid, commonly the value of the engagement or a year's fees, is a standard request and most procurement teams have an equivalent version in their own supplier terms. Raise it early in the negotiation, not after the rest of the contract has otherwise been agreed, so it does not read as an afterthought.
Is cyber liability insurance necessary if I'm not storing client data myself?
If you ever touch client data, even briefly, to build, test or evaluate a model, you carry some exposure, and cyber liability cover responds to the costs of a breach, including notification and remediation. Ask a broker to assess your actual data handling rather than assuming your involvement is too limited to matter.
Where the figures come from
Every rate and salary quoted in this article is a median or percentile of figures advertised in UK job postings over the six months to 8 August 2026. They are not rates paid, and the gap widens at the top of a range.
- IT Jobs Watch, UK contract rates, 6 months to 8 August 2026, read 2026-08-08.
- IT Jobs Watch, UK permanent salaries, 6 months to 8 August 2026, read 2026-08-08.