Terms of Trade for AI Software Companies in New Zealand

If you run an AI software business in New Zealand, your terms of trade do more than set payment dates. They decide who carries the risk when a model gives a bad output, whether you can use customer data to improve your product, and what happens if a client claims your tool caused them loss. Founders often make three expensive mistakes here. First, they rely on generic software terms that do not deal properly with AI-generated outputs. Second, they promise too much in sales conversations, then sign a contract that does not match those promises. Third, they overlook New Zealand rules around fair dealing, privacy and service quality.

The result is usually avoidable conflict about scope, liability, data use and termination. This guide explains what terms of trade for AI software company arrangements should cover in New Zealand, the legal issues to check before you sign, and the mistakes that commonly catch AI founders and SMEs accepting AI vendor terms.

Overview

Terms of trade for an AI software company set the commercial and legal rules for how your software is supplied, used and supported. In New Zealand, they should be tailored to AI-specific risks, not copied from a standard SaaS template without changes.

Well-drafted terms usually deal with both ordinary software issues and the extra uncertainty that comes with machine learning tools, automated outputs, data processing and customer reliance on results.

• Define exactly what the AI software does, and what it does not do
• Set payment terms, renewal rules, suspension rights and price change mechanisms
• Explain who owns the platform, the customer data, inputs, outputs and any derived learning
• Deal clearly with privacy, confidentiality and permitted data use
• Limit liability for inaccurate outputs, downtime, third party tools and misuse
• State any acceptable use restrictions, especially for high-risk or unlawful use cases
• Set support, service levels, change management and termination rights
• Make sure your wording aligns with the Fair Trading Act, the Consumer Guarantees Act where relevant, and other New Zealand legal obligations

What Terms of Trade for AI Software Company Means For New Zealand Businesses

For New Zealand businesses, terms of trade for AI software company arrangements are the contract rules that shape the relationship between the provider and the customer. They matter most at the exact moment expectations diverge, especially when a founder says one thing in a demo and the written terms say another.

If you supply AI tools, your terms of trade are usually the supplier's standard customer contract. If you buy AI software, they are often the vendor's standard terms that you are asked to accept before you sign. In both cases, the core question is the same: who is responsible when the product underperforms, produces a flawed output, or handles data in a way the customer did not expect?

Why AI terms need more than a standard software contract

Normal software terms deal with licences, fees, access rights and service availability. AI products add a different layer of risk because outputs can be probabilistic, influenced by training data, and unsuitable for some high-stakes decisions.

This is where founders often get caught. A sales team may position the product as highly accurate or "decision-ready", but the contract quietly says outputs are provided on an "as is" basis and must not be relied on without human review. If those two messages do not line up, you create room for a dispute and possible issues under the Fair Trading Act 1986 if the marketing overstates capability.

Common forms these terms take

The legal document may be called terms of trade, terms and conditions, master services agreement, software subscription agreement, SaaS agreement, order form terms, enterprise licence agreement, or customer terms. The label matters less than the substance.

A New Zealand AI software contract commonly covers:

• the subscription or licence granted to the customer
• implementation or onboarding services
• usage limits, seat limits or API call limits
• customer obligations around lawful and appropriate use
• data handling and privacy responsibilities
• warranties, disclaimers and risk allocation
• termination, transition and data return or deletion

What AI software founders should address specifically

The biggest drafting gap is usually around data and outputs. Your terms should say what happens to prompts, uploaded files, customer datasets, generated text, recommendations, reports and model improvements.

If your product learns from usage data, say so clearly. If your product does not use customer content for training, say that clearly too. Silence creates uncertainty, and uncertainty often pushes enterprise customers to demand heavier custom terms later.

You should also be precise about intended use. For example, an AI recruiting tool, health-support tool, fraud detection engine or legal document assistant may each need different restrictions, warnings and review obligations. A clause that simply says "customer must comply with all laws" is not enough if the product is marketed into higher-risk decision areas.

New Zealand legal context

Several New Zealand legal rules can affect how these contracts work in practice. The Fair Trading Act restricts misleading or deceptive conduct in trade, so your pre-contract statements, website copy, demos and proposal documents should match the legal terms. A liability cap will not necessarily save a business from trouble if the customer says they were misled before signing.

The Consumer Guarantees Act 1993 may also matter in some situations. Many AI software deals are business-to-business, and businesses often contract out of the Act where the law allows and the arrangement is in trade. That needs proper written terms, and it will not always be available or appropriate. The position depends on the customer and the deal.

If personal information is involved, the Privacy Act 2020 sits in the background as well. Your contract should help clarify each party's role, especially where the software provider stores, analyses or processes identifiable information on the customer's behalf. That is separate from having a public privacy notice, but the two should align.

Legal Issues To Check Before You Sign

Before you sign a contract for AI software, the main legal job is to test whether the written terms match the real product, the sales pitch and the data flows. If they do not match, the contract will not protect the relationship when something goes wrong.

Scope, functionality and performance claims

The contract should describe the product in a way that is specific enough to avoid argument later. Vague wording such as "AI business optimisation tools" leaves too much room for assumptions.

Look for detail on:

• what features are included in the current subscription
• whether beta or experimental features are covered differently
• what level of uptime or availability is promised, if any
• whether outputs are advisory only or suitable for operational use
• whether human review is required before action is taken

If the provider promised a feature in a demo, get it into the agreement or order form. Before you rely on a verbal promise, ask where it appears in the contract documents.

Data ownership, use and training rights

Data clauses matter more in AI contracts than many founders expect. The customer will usually want to retain ownership of its source data and uploaded content. The provider will usually want rights to host, process and use that data to operate the service.

The contract should separate at least four categories:

• the provider's pre-existing software, models and intellectual property
• the customer's input data, prompts and uploaded materials
• the outputs generated by the service
• analytics, learnings or de-identified usage information created from operation of the platform

Each category can have a different ownership and licence position. If the provider wants to use customer data for model training, benchmarking or product improvement, that should be express and easy to understand. If the customer prohibits that use, the contract should also say so directly.

Privacy and confidential information

If personal information is processed through the tool, the contract should say what protections apply and who is responsible for what. In many B2B deals, the customer controls the collection purpose and the provider processes information under the customer's instructions, but that is not always the full picture.

Check whether the agreement deals with:

• security measures and access controls
• subprocessors or third party hosting providers
• cross-border data storage or access
• notification obligations for privacy or security incidents
• retention, deletion and export of customer data on exit

Confidentiality wording should cover customer data, technical information, pricing and any non-public model or system information shared between the parties. The provider will often try to carve out aggregated or de-identified data. That may be reasonable, but the wording needs care.

Liability, indemnities and exclusions

The liability section is usually where the financial risk gets allocated. AI vendors often try to exclude almost everything, especially indirect loss, lost profits, data loss and reliance on outputs. Customers, particularly SMEs using the software in core operations, should look closely before they accept that position.

The practical questions are:

• what losses are excluded altogether
• what the liability cap is, and whether it is tied to fees paid over a period
• whether some claims are carved out of the cap, such as confidentiality breaches or IP infringement
• whether the customer must indemnify the provider for misuse, unlawful content or prohibited use cases
• whether the provider gives any IP infringement indemnity if the software itself causes a third party claim

A cap based on one month of fees may be far too low if the software is embedded deeply into operations. On the other side, providers need to make sure they are not accepting unlimited exposure for every flawed output a customer chooses to rely on.

Acceptable use and high-risk use cases

AI products should have tailored use restrictions. This matters if the tool could be used for employment screening, credit decisions, healthcare support, legal advice, safety-critical systems, or content that breaches the law or third party rights.

Good terms usually restrict:

• unlawful, harmful or misleading use
• use that infringes intellectual property or privacy rights
• automated decision making without required human oversight
• attempts to reverse engineer, scrape or abuse the platform
• feeding in sensitive information where the product is not designed for it

These clauses do not replace operational compliance, but they help set boundaries and reduce foreseeable misuse.

Term, renewal and exit rights

Subscription software disputes often arise at renewal or termination. The contract should state the initial term, how renewals happen, what notice is required, and what each party can do if the other breaches.

Before you accept the provider's standard terms, check:

• whether the contract auto-renews
• whether fees can increase mid-term or only on renewal
• whether there is a right to suspend for non-payment or security concerns
• what happens to stored data after termination
• whether there is any transition support or export right

If the customer would struggle to move away quickly, exit planning is not a side issue. It is part of the commercial risk.

Common Mistakes With Terms of Trade for AI Software Company

The most common mistake is treating AI terms like ordinary software boilerplate. That approach usually leaves gaps around outputs, data rights and customer reliance, which are exactly the points that matter when a disagreement starts.

Using a generic template without AI-specific edits

A standard SaaS template can be a useful base, but only if it is revised for the actual product. If your software generates content, recommendations or classifications, you need clear wording on accuracy, review obligations and prohibited reliance.

Many templates also assume the provider does not learn from customer activity. That assumption can be wrong for AI systems and can create direct conflict with your actual technical processes.

Letting sales language outrun the contract

Founders often focus on closing the deal and worry about the paperwork later. The problem is that statements in decks, emails and demos can shape what the customer says they were promised.

If you describe the product as fully automated, highly accurate, compliant by design, or suitable for sensitive decisions, your written terms need to support those claims or qualify them properly. The Fair Trading Act risk sits here, not just in the fine print.

Ignoring who can rely on outputs

AI output is often used by teams beyond the immediate buyer. A customer may circulate reports internally, use generated content in marketing, or plug recommendations into downstream systems.

Your terms should address whether outputs are for internal business use only, whether they can be shared with affiliates or advisers, and whether the provider accepts any responsibility to third parties. If that is left open, the scope of risk can expand fast.

Missing the operational realities of support and change

AI software changes quickly. Models are updated, interfaces shift and features are retired. If the provider has broad rights to change the service at any time, the customer may end up paying for something materially different from what was bought.

The contract should say enough about support, planned changes, deprecated features and notice periods. This is particularly important where the software is integrated into workflows or customer-facing systems.

Forgetting downstream legal documents

Terms of trade do not sit alone. They should align with the rest of the business's legal documents and processes.

Depending on the product, that may include:

• statements made in proposals, scopes and order forms
• privacy disclosures and internal privacy procedures
• acceptable use or platform rules
• contractor and employee confidentiality obligations
• intellectual property assignments from developers and consultants

If these documents pull in different directions, the main contract becomes harder to enforce cleanly.

Accepting one-sided enterprise paper without checking the commercial effect

SMEs buying AI software sometimes assume the vendor's standard enterprise terms are non-negotiable. That is not always true. Even where the core position will not change, some targeted edits can materially reduce risk.

For example, a customer might ask for better data deletion wording, clearer service descriptions, a more realistic liability cap, or notice before material changes. A provider might need stronger wording around prohibited uses, customer data rights or non-payment suspension. These are not abstract legal points. They affect how the deal works in day-to-day business.

FAQs

Do AI software companies in New Zealand need special terms of trade?

Usually, yes. Standard software terms often miss AI-specific issues such as output accuracy, human review, training rights, model changes and limits on reliance. Tailored wording is usually worth it.

Can an AI software provider exclude all liability for bad outputs?

Not automatically. Providers often try to limit liability heavily, but the enforceability and commercial fairness of that approach depends on the wording, the deal context and any pre-contract representations. New Zealand fair trading rules also matter.

Who owns AI-generated outputs under a software contract?

The contract should say. Some agreements give the customer rights to use outputs, while the provider keeps ownership of the underlying platform and models. Ownership and licence terms should be drafted clearly, especially if outputs are commercially valuable.

Should the contract mention privacy if the software handles business data?

Yes, especially where personal information may be included. The agreement should address processing rights, security, incident response, sub-processors, storage location and deletion or return of data at the end of the contract.

Can a New Zealand business contract out of the Consumer Guarantees Act in an AI software deal?

Sometimes, in business-to-business dealings where the legal requirements are met and both parties are in trade. The wording needs to be correct, and it will not be suitable in every situation.

Key Takeaways

• Terms of trade for AI software company arrangements should be drafted for AI-specific risks, not lifted from a generic software template without changes.
• The contract should clearly cover scope, functionality, data rights, outputs, privacy, confidentiality, acceptable use, liability and exit arrangements.
• Marketing claims, demos and sales promises should match the written terms, especially under the Fair Trading Act.
• Data ownership and data use clauses need special attention, including any rights to use customer inputs or usage data for model training or product improvement.
• Customers and providers should both review liability caps, exclusions and indemnities carefully before they sign.
• High-risk use cases, human review requirements and limits on reliance should be stated plainly in the agreement.
• Privacy Act issues, security responsibilities and data deletion or export rights should be addressed where personal information or sensitive business data is involved.
• If you are reviewing or negotiating terms of trade for AI software company and want help with contract drafting, liability clauses, data use terms, privacy obligations, you can reach us on 0800 002 184 or team@sprintlaw.co.nz for a free, no-obligations chat.