Disclaimers and Liability Limits for AI Software Companies in New Zealand

AI software companies in New Zealand often know they need legal protection, but many get the basics wrong. A common mistake is pasting in a generic overseas disclaimer that does not fit New Zealand law. Another is promising too much in sales conversations, then trying to claw risk back in the contract. A third is setting a liability cap that looks tidy on paper but will not hold up against the way the product is actually sold, supported, and used.

If you supply AI tools, data-driven software, automation products, or embedded AI features, your disclaimers and liability limits are not just legal fine print. They shape what risk you keep, what risk you shift, and how disputes are likely to play out if the software makes a bad recommendation, produces inaccurate output, causes a downstream loss, or mishandles data. This guide explains what disclaimers and liability limits for AI software company arrangements usually cover in New Zealand, what to watch before you sign, and where founders often get caught.

Overview

Disclaimers and liability clauses help define the boundaries of responsibility when AI software is used by customers, teams, and end users. In New Zealand, those clauses need to work with contract law, the Fair Trading Act, the Consumer Guarantees Act where relevant, and privacy obligations, rather than trying to sidestep them.

A strong position usually comes from matching your contract wording to how the product is actually marketed, implemented, and supported. If your sales pitch says one thing and your limitation clause says another, the contract may not protect you in the way you expect.

• State clearly what the AI product does, and what it does not do
• Separate factual warranties from marketing language and future-looking statements
• Set fair, tailored liability caps rather than copying broad exclusions from foreign templates
• Deal expressly with inaccurate outputs, customer data, third party tools, and service interruptions
• Check whether the customer is a business or consumer, because that changes what can be contracted out of
• Make sure privacy, security, acceptable use, and indemnity clauses align with the liability framework

What Disclaimers Liability Limits for AI Software Company Means For New Zealand Businesses

For New Zealand AI businesses, disclaimers and liability limits are the contract terms that define the edge of your responsibility. They tell the customer what reliance is reasonable, what losses you will and will not cover, and how much exposure your business is taking on if something goes wrong.

That matters because AI products create a familiar problem in a new form. Customers may rely on outputs for recruitment, credit decisions, forecasting, coding, compliance tasks, customer service, document review, health-adjacent analysis, or operational workflows. If the tool produces a false result or makes an unsafe suggestion, the customer may say your product caused the loss. Your written terms need to address that risk directly.

What a disclaimer usually does

A disclaimer is not a magic shield. Its real job is to clarify the limits of the product, the assumptions behind it, and the kinds of reliance that are not permitted.

For an AI software company, disclaimers often deal with:

• accuracy of outputs
• bias, hallucinations, incomplete results, or model drift
• customer responsibility to review and validate output
• third party datasets, APIs, cloud providers, or integrations
• service availability and changes to features
• whether the product is advisory only and not professional advice

For example, if your product generates business recommendations, your agreement might say the output is provided for general informational use and should not be treated as legal, financial, medical, or other regulated advice. If your product supports internal decision making, you may also state that the customer remains responsible for final decisions, approvals, and compliance checks.

What a liability limit usually does

A liability limit sets the financial and legal boundaries of a claim. It usually answers three practical questions: what types of loss are excluded, what claims are carved out, and what is the maximum amount payable if liability does arise.

Common examples include:

• excluding indirect or consequential loss, such as lost profits, lost opportunity, or reputational harm
• capping total liability to fees paid over a set period
• carving out certain matters from the cap, such as fraud, wilful misconduct, breach of confidentiality, or some privacy breaches
• limiting claims to a certain time period after the event

The right cap depends on the product and customer profile. A low-cost self-serve AI tool sold to small businesses may justify one approach. Enterprise software used in a sensitive workflow may need a different structure, with separate caps for service issues, data loss, and confidentiality breaches.

Why New Zealand law changes the drafting

New Zealand businesses cannot simply write “use at your own risk” and expect that to settle the issue. The enforceability of your disclaimers depends on the customer relationship, how the terms are presented, and whether any law limits your ability to contract out of statutory protections.

If you are supplying to consumers, the Consumer Guarantees Act may apply and you generally cannot contract out of it. If you are dealing business to business, contracting out may be possible in some cases, but only if the statutory requirements are met and the wording is properly handled. The Fair Trading Act also matters, because your marketing and sales claims must not be misleading or deceptive. A disclaimer will not necessarily rescue a statement that overpromised in a demo, proposal, or onboarding call.

Privacy law matters too. If your AI tool handles personal information, your liability clauses need to sit alongside your Privacy Act obligations, your privacy notice, and your actual data handling practices. A broad disclaimer about data risk is unlikely to help if your business has promised certain security standards or failed to follow its own process.

Founder moments where this becomes real

This issue usually surfaces at predictable moments:

• before you accept the provider's standard terms from a cloud or model supplier and pass those risks downstream to your own customer
• before you sign a large customer contract with procurement changes to your liability cap
• before you rely on a verbal promise made by sales staff about accuracy, uptime, or suitability
• before you offer a pilot, proof of concept, or discounted rollout without adjusting the legal terms
• before you deploy AI into a regulated or high impact use case

This is where founders often get caught. The legal wording may look settled, but the commercial team may have promised custom outcomes, human-level accuracy, or risk-free automation. If that mismatch is not fixed before you sign, the contract can become much less protective.

Legal Issues To Check Before You Sign

Before you sign a customer contract, the main job is to make sure the risk allocation matches the actual product, the customer use case, and the laws you cannot contract around. Good contract drafting is specific, consistent, and realistic.

1. Product description and scope of use

Your disclaimer framework starts with a clear description of what the software is for. If your agreement is vague, customers may argue they were entitled to rely on the system for more than you intended.

The contract should identify:

• the core functionality of the AI feature or platform
• whether outputs are deterministic, predictive, generative, or assistive
• any stated performance limits
• the customer's responsibility for review, approval, and implementation
• restricted or prohibited use cases

If certain uses create higher risk, such as employment screening, credit assessment, safety-critical tasks, or legal advice workflows, say so plainly. Do not leave that to a support article or sales slide.

2. Warranties and performance promises

Every promise in your contract should be intentional. The more specific the warranty, the harder it may be to limit liability if the promise is not met.

Many AI software companies give limited warranties about access, support, and conformity with documentation, while avoiding guarantees about uninterrupted service, absolute accuracy, or fitness for every customer purpose. That approach can be sensible, but only if your marketing team is not making stronger claims elsewhere.

Check for inconsistency across:

• the master services agreement or software terms
• the order form or proposal
• product documentation
• sales emails and demos
• service level commitments

If a customer has relied on a detailed statement from your team, that may become part of the dispute even if the liability clause says otherwise.

3. Liability caps and exclusions

A liability cap should be chosen, not guessed. The number needs to reflect contract value, the risk of downstream harm, your insurance position, and the role your product plays in the customer's operations.

Points to settle before you sign include:

• whether the cap is based on fees paid in the last 12 months, the full contract value, or a fixed dollar amount
• whether different claims should have different caps
• whether indirect and consequential loss is excluded
• what exceptions sit outside the cap
• whether the cap applies per claim or in aggregate

In practice, customers often push for carve-outs relating to confidentiality, privacy, intellectual property infringement, and wilful misconduct. Suppliers often push back on unlimited liability. The final position should be commercially sensible and internally consistent.

4. Consumer law and fair dealing rules

You cannot assess disclaimers in isolation from New Zealand consumer law. If the customer is a consumer, statutory guarantees may override broad contractual exclusions. If the deal is business to business, any contracting-out wording needs to be carefully drafted and appropriate to the relationship.

The Fair Trading Act is just as important. If your advertising says the tool is “error-free”, “fully compliant”, or “safe to rely on without review”, that can create trouble even if the contract contains a careful limitation clause. The contract should support your real-world messaging, not contradict it.

5. Privacy, data use, and security allocation

If your AI software processes personal information, privacy risk needs its own attention. A liability clause that ignores data handling issues is often incomplete.

Make sure the agreement covers:

• who controls and who processes personal information
• what data can be used for model training, service improvement, or analytics
• whether customer data is de-identified
• security commitments and incident notification expectations
• cross-border service providers and sub-processors

Customers may accept a normal service outage cap but demand stronger treatment for unauthorised disclosure or misuse of data. That is a commercial risk question as much as a legal drafting question.

6. Intellectual property and third party components

AI products often rely on third party models, datasets, or software layers. Your contract should say what IP rights the customer receives, what restrictions apply, and how third party dependencies affect liability.

Founders should also think carefully about infringement claims relating to training data, generated content, or user inputs. Some businesses offer a limited IP indemnity, some exclude generated outputs, and some pass through the limits of upstream providers. None of those choices are automatically right. The key is making the allocation explicit before you sign.

7. Indemnities, claims process, and remedies

Indemnities can quietly undo a carefully drafted liability cap. If you agree to broad indemnity wording for all loss arising out of the software, the cap may not protect you in the way you intended.

Review:

• whether the indemnity is tied to specific risks only
• whether it is capped or uncapped
• who controls defence and settlement of claims
• whether service credits, re-performance, or termination rights are the main remedy for service failures

This is one of the most important checks before you accept the provider's standard terms from a larger customer or enterprise procurement team.

Common Mistakes With Disclaimers Liability Limits for AI Software Company

The most common mistake is treating the disclaimer as a document problem when it is really a product, sales, and operations problem. If the business behaves as though the tool can be trusted without limits, the contract may not save you.

Copying overseas templates

US and UK templates often use concepts, statutory references, and market conventions that do not cleanly fit New Zealand. Even where the wording looks familiar, it may not deal properly with the Fair Trading Act, the Consumer Guarantees Act, or the way your business actually contracts.

Founders often spend money on setup, product development, and sales before checking whether the legal terms are fit for local use. That usually leads to rushed edits at the point of a major customer deal.

Overstating what the AI can do

If the product is still probabilistic, experimental, or dependent on user review, your sales language should say that. Founders sometimes describe the system as replacing expert judgement, then try to soften that promise in the terms. That mismatch is a major risk.

Problem phrases often include:

• fully automated compliance
• guaranteed accurate recommendations
• bias-free decision making
• human-equivalent analysis
• set and forget workflows

Those claims can create expectations that are hard to unwind later.

Setting an arbitrary cap

A cap of three months' fees or twelve months' fees is common, but not automatically suitable. If your software affects large commercial decisions, sensitive datasets, or customer-facing outputs, the customer may see a very low cap as unrealistic. On the other hand, agreeing to a cap that far exceeds contract value may expose a startup to existential risk.

The better approach is to assess the use case, likely loss profile, bargaining position, and available insurance, then set a cap you can defend commercially.

Ignoring pilots and free trials

AI companies often run pilots before a paid rollout. That is exactly when customers may test the tool in live workflows and make important decisions based on output. If your pilot terms are silent on reliance, liability, confidentiality, and data use, the trial can create risk long before the full contract is signed.

Even a short evaluation arrangement should address:

• permitted use
• no production reliance, if that is your position
• data handling rules
• feedback rights
• liability limits during the trial period

Forgetting the customer's own obligations

Suppliers often focus on excluding their own liability but forget to impose positive obligations on the customer. If the customer must validate outputs, maintain human review, use the software lawfully, or avoid prohibited inputs, say so directly.

That matters because many disputes turn on mixed responsibility. If the customer ignored instructions, uploaded problematic data, or used the product outside scope, your agreement should make that relevant.

Leaving privacy and AI terms disconnected

Many businesses have a privacy statement prepared separately from their customer contract. If those documents conflict on data use, retention, training rights, or security commitments, the confusion can create both legal and commercial trouble.

The main risk is not just regulatory. Customers may argue they were promised one thing in a policy document and another in the agreement.

FAQs

Can an AI software company fully exclude liability in New Zealand?

No. You can limit and allocate risk by contract, but some legal obligations cannot simply be excluded, especially in consumer contexts. Even in business to business contracts, broad exclusions still need to be reasonable in context and consistent with the rest of the deal.

Do disclaimers protect us if the AI gives inaccurate output?

Sometimes, but only if the disclaimer is clear, properly incorporated into the contract, and consistent with how the software was marketed and used. A disclaimer is much more effective when the agreement also requires the customer to review and validate outputs.

Should our liability cap be tied to subscription fees?

Often yes, but not always. A fees-based cap is common for SaaS and AI products, though some deals justify different caps for confidentiality, privacy, or IP issues. The right structure depends on the contract value and the type of harm that could realistically arise.

Do we need separate terms for pilots or proofs of concept?

Usually yes. Trial use creates its own risks, especially where customers may test real data or rely on outputs in live settings. Short-form pilot terms can prevent arguments later about reliance, data rights, confidentiality, and liability.

Can we rely on a website disclaimer alone?

Usually not for higher value or business-critical supply arrangements. A website disclaimer may help with general messaging, but your main protections should sit in the signed contract, accepted online terms, order form, and related documents that clearly govern the relationship.

Key Takeaways

• Disclaimers and liability limits for AI software company contracts should match the actual product, customer use case, and sales process
• New Zealand law matters, especially the Fair Trading Act, the Consumer Guarantees Act where relevant, contract law, and privacy obligations
• A disclaimer does not fix overpromising in demos, proposals, or onboarding conversations
• Liability caps should be tailored to the risk profile, not copied from a foreign template or chosen at random
• Privacy, IP, third party tools, indemnities, and pilot arrangements need to align with the overall liability framework
• Founders should review these issues before they sign, before they accept the provider's standard terms, and before they rely on a verbal promise

If you want help with customer contracts, contract review, privacy terms, liability caps, and pilot agreements, you can reach us on 0800 002 184 or team@sprintlaw.co.nz for a free, no-obligations chat.