Alex is a co-founder of Sprintlaw. Alex previously worked at a top-tier firm as a lawyer specialising in technology and media contracts, and founded a digital agency which he sold in 2015.
Priya runs a small AI startup in Wellington. Her team has built a tool that drafts customer support replies, and three beta customers are ready to test it next week. Everyone is excited, until one customer asks, "Can you send your SaaS terms, AI policy and privacy wording before we upload any data?"
That question can feel like a speed bump, but it's actually a good sign. If a customer is about to rely on your AI tool, send you data or invite their team into your platform, they will want to know what they're agreeing to. Clear AI SaaS terms help you answer those questions once, rather than negotiating the basics every time someone wants a trial.
For an AI SaaS business, ordinary software terms are a starting point, not the whole picture. You'll also need to deal with inputs, outputs, model limitations, data use, privacy, training rights, human review and risk allocation. The goal isn't to scare customers with legal language. It's to set practical rules so both sides know how the product can be used, what it doesn't do and who is responsible if something goes wrong.
Why Your AI SaaS Terms Matter Before Launch
Your terms are the contract between your business and your customers. They explain what you provide, what the customer is allowed to do, what they must pay, what happens to their data and how the relationship can end.
For a SaaS product, customers often accept the terms online by ticking a box, signing an order form or continuing to use the platform after seeing a link to the terms. That means your documents need to be ready before the beta, pilot or public launch, not after you have already onboarded users.
This is especially important for AI products because customers may use the tool in ways that create real world consequences. For example, your tool might draft customer support replies, rank sales leads, summarise HR documents, create marketing copy or suggest next steps for an internal process. Even if the tool is only assisting a person, the customer needs to understand that outputs may be inaccurate, incomplete or unsuitable for some decisions.
Good AI SaaS terms also make sales smoother. Larger customers may ask for your privacy position, service levels, security standards and AI output rules before they approve procurement. If your terms already cover those points clearly, you can often avoid long back and forth negotiations.
What Your AI SaaS Terms Should Include
Your SaaS terms should cover the commercial and operational basics first. That includes who can use the platform, what subscription plan applies, how fees are calculated, how billing works, what support is included and how either party can terminate.
You should also describe the licence you're giving the customer. In most cases, the customer isn't buying your software. They're getting a limited right to access and use your platform during the subscription period, usually for their internal business purposes. Your terms should say whether they can add users, share logins, integrate with other systems or use the platform for clients.
It's also important to include acceptable use rules. These might prohibit users from reverse engineering the platform, attempting to extract the model or source code, uploading unlawful or harmful content, infringing another person's rights, bypassing usage limits or using the platform to build a competing product. For AI tools, you may also want to prohibit customers from using the tool to generate misleading content, spam, discriminatory decisions, illegal advice or content that breaches applicable laws.
If you offer a beta or free trial, deal with that separately. A beta may have bugs, changing features, limited support and lower availability. Your terms should make it clear whether beta use is free or paid, whether data may be deleted at the end of the beta and whether you can change or withdraw beta features at short notice.
Data, Privacy And Confidentiality
AI SaaS products often work because customers upload useful information. That may include customer enquiries, documents, chat transcripts, employee information, sales data or confidential business material. Your terms need to explain how you handle that information.
If you collect or handle personal information in New Zealand, you need to think about the Privacy Act 2020. Your privacy wording should tell customers and users what personal information you collect, why you collect it, who you share it with, where it may be stored, how people can access or correct it and how they can contact you about privacy issues.
Your SaaS terms should sit neatly with your Privacy Policy. The terms can cover the contractual side, such as the customer's promise that they have the right to upload data into the platform. The Privacy Policy can cover the public facing explanation of your privacy practices. If you use hosting providers, analytics tools, payment processors or AI infrastructure providers, you should understand what those providers do with the data and whether data is stored or accessed outside New Zealand.
Confidentiality is also key. Customers may upload commercially sensitive data, and you may share technical information with them during onboarding. Your terms should require both sides to protect confidential information and only use it for the purposes of the SaaS relationship. You can also set out sensible exceptions, such as information that's already public or must be disclosed by law.
Finally, include a process for security incidents and notifiable privacy breaches. You don't need to promise perfect security, but you should explain that you'll use reasonable safeguards and will notify affected customers where required or appropriate.
AI Outputs, Accuracy And Human Review
This is where AI SaaS terms need extra care. Your product may generate text, recommendations, summaries, predictions, classifications or other outputs. Customers should understand that these outputs are generated by software and may not always be accurate, complete, current or appropriate for their situation.
Your terms should explain what the customer is responsible for checking. For example, if your tool drafts customer support replies, the customer should review replies before sending them. If your tool summarises documents, the customer should check the source material before relying on the summary. If your tool helps with compliance, finance, employment or health related workflows, the customer shouldn't treat the output as a substitute for professional advice or their own decision making.
It's also worth being specific about prohibited reliance. If your AI tool isn't designed for high risk decisions, say so. You might restrict use for decisions about a person's employment, credit, insurance, housing, health, legal rights or access to essential services unless you have specifically agreed to that use and built the right safeguards.
Accuracy wording should be balanced. You don't want to overpromise, but you also shouldn't make your product sound useless. A practical approach is to say that you'll provide the platform with reasonable care and skill, but AI outputs may vary and the customer remains responsible for how they use, review and act on them.
Intellectual Property, Inputs And Training Data
Intellectual property is often one of the biggest questions in AI SaaS terms. Your customer may ask, "Do we own what we upload? Do we own what the AI generates? Will you train your model on our data?" Your terms should answer those questions plainly.
Usually, you'll want to confirm that the customer owns or retains rights in their inputs, such as prompts, documents and data they upload. You'll also want to confirm that you own your platform, software, models, workflows, prompts, know how, templates, documentation and improvements.
For outputs, there are different approaches. Some SaaS providers let customers own or use outputs for their business purposes, subject to the terms. Others grant a broad licence rather than transferring ownership, particularly where outputs may be similar for multiple users. Whatever approach you take, be clear. Customers need to know whether they can publish, edit, sell, share or rely on outputs.
Training data deserves its own wording. If you want to use customer data to train or improve your models, say that clearly and get the right permission. Many business customers won't accept their confidential data being used to train models for other customers. You might offer an opt in approach, use de identified or aggregated data, or promise that customer content won't be used for model training without written consent.
If you rely on a third party AI provider, check its terms too. Your promises to customers must match what your upstream provider allows. For example, if the provider can store prompts for abuse monitoring, or restricts certain categories of use, your own terms shouldn't promise something inconsistent.
Subscriptions, Support And Service Levels
Your commercial terms should be easy to understand. Set out the fees, billing cycle, GST treatment, payment method, renewal process and what happens if a payment is late. If pricing depends on users, seats, credits, tokens, API calls or data volume, explain how usage is measured and when overage fees apply.
Customers will also want to know what level of service to expect. You can include support hours, support channels, response targets and any uptime commitments. If you're an early stage startup, be careful about promising enterprise grade service levels before you can consistently meet them.
You should reserve the right to maintain and improve the platform. SaaS products change quickly, and AI products change even faster. Your terms can allow you to release updates, change features, improve models and make reasonable changes to the service, while still giving customers notice of material changes where appropriate.
Termination rights are also important. You may need to suspend or terminate access if a customer doesn't pay, breaches acceptable use rules, creates security risks or uses the product unlawfully. Customers may want the right to cancel at the end of a subscription period and export their data. Your terms should explain what happens to data after termination, including any retention period and deletion process.
Liability, Consumer Law And Fair Terms
Every SaaS business needs to think carefully about risk. AI tools can be powerful, but they can also produce unexpected results. Your terms should set sensible limits on your liability, while staying within New Zealand law.
Common protections include excluding liability for indirect loss, loss of profits, loss of revenue, loss of data and business interruption, and capping liability at an amount linked to fees paid over a recent period. You might also include indemnities where the customer misuses the platform, uploads unlawful content or breaches someone else's rights.
However, you can't simply contract out of every legal responsibility. New Zealand laws such as the Fair Trading Act 1986 and Consumer Guarantees Act 1993 may apply depending on who your customers are and how the product is supplied. If you deal with consumers, you need to be careful not to exclude rights that can't be excluded. If you deal with business customers, you may be able to include certain exclusions or limitations, but the wording needs to be clear and appropriate.
Unfair contract terms are another issue. Standard form consumer contracts and many small trade contracts can be reviewed under the Fair Trading Act. Terms that are heavily one sided, not reasonably necessary to protect your legitimate interests or would cause detriment may be at risk. This doesn't mean you can't protect your business. It means your terms should be fair, transparent and proportionate.
For AI SaaS, the best risk wording is practical. Tell customers what the product is and isn't designed to do. Require human review where needed. Set acceptable use rules. Be clear about data responsibilities. Then use liability limits to manage the remaining risk.
Key Takeaways
- AI SaaS terms should be ready before your beta, pilot or public launch, especially if customers will upload data or rely on outputs.
- Your terms should cover access rights, acceptable use, pricing, support, service changes, suspension and termination.
- Be clear about data handling, privacy, confidentiality, overseas providers and what happens if there's a security incident.
- AI specific wording should explain output limitations, customer review obligations and any restrictions on high risk use.
- Set out who owns inputs, outputs, your platform and improvements, and whether customer data can be used for model training.
- Liability limits should be fair, transparent and consistent with New Zealand consumer, privacy and fair trading laws.
If you'd like help with AI SaaS terms, privacy wording, model training clauses or launch contracts, you can reach us at 0800 002 184 or team@sprintlaw.co.nz for a free, no-obligations chat.
Drafted by Taylor, Sprintlaw's AI, and overseen by Alex Solo. How we write
Get your customer-facing terms right
What should your privacy and online terms cover?
If you collect customer data, sell online or run marketing campaigns, your public terms and privacy documents should match the real customer journey.







