IP Ownership Issues for AI Product Startups in New Zealand

If you are building an AI product in New Zealand, one of the fastest ways to lose value is to assume your company automatically owns everything your team creates. That is often wrong.

Founders regularly make the same mistakes: they use contractors without written IP assignment clauses, they build on third party models or datasets without checking licence terms, or they let a co-founder code the product before the company is properly set up and documents are signed.

The problem is not just theoretical. Investors, acquirers, enterprise customers and commercial partners will usually ask who owns the code, model improvements, data outputs, brand and training materials. If the answer is vague, the deal can slow down or fall over. This guide explains how IP ownership works for AI product startups in New Zealand, when the issue comes up in real founder decisions, what documents usually matter, and where businesses commonly get caught before they sign, launch online, or invest in branding.

Overview

For New Zealand AI startups, IP ownership is usually split across several layers, not just the app code. Your company may need rights in software, data, prompts, model outputs, trade marks, confidential information and contracts with people who helped create the product.

The main legal question is simple: can the business clearly prove it owns, or has permission to use, every core asset needed to operate and commercialise the product?

• Confirm who created the code, workflows, prompts, datasets, documentation and brand assets.
• Check whether employees, founders and contractors have signed clear IP ownership and confidentiality terms.
• Review any third party AI tools, models, open source components and datasets for licence limits and ownership restrictions.
• Decide what should be protected as copyright, trade marks, confidential information or contract rights.
• Make sure customer terms explain who owns inputs, outputs, feedback and product improvements.
• Sort out privacy and data use permissions if personal information is used to train, test or improve the product.

What IP Ownership AI Product Startups Means For New Zealand Businesses

IP ownership for an AI startup means identifying every valuable intangible asset in the product and making sure the business, not just an individual, has legal control over it.

Many founders think of intellectual property as just copyright in software. In practice, AI products often combine several different rights and risk areas.

What counts as IP in an AI product?

An AI startup may have protectable rights in the product stack itself, as well as contractual rights over how others use it. Depending on your product, that can include:

• source code, object code and technical architecture
• training pipelines, prompt structures and workflow designs
• UI copy, product documentation and internal playbooks
• brand names, logos and taglines that may be protected through trade mark registration
• proprietary datasets, labelled data and data schemas
• customer lists, pricing models and go to market strategy kept as confidential information
• rights to use third party models, APIs, libraries and hosted infrastructure
• customer contracts that allocate rights in outputs, feedback and derivative works

Some of these rights arise automatically, such as copyright in original works. Others depend heavily on registration or contract. A trade mark, for example, is much easier to enforce if you apply to register it in New Zealand before you invest in branding.

Does the company automatically own founder-created IP?

No, not always. This is where founders often get caught.

If a founder built the prototype before the company existed, the founder may personally own that IP unless it has been properly assigned to the company. The same issue comes up where a founder used an old side project, code from a previous job, or material developed through a university or funded research arrangement.

Before you spend money on company setup, raise capital or sign a shareholder deal, make sure any pre-incorporation IP is transferred to the company in writing. If your business structure changes later, check whether the transfer still reflects the actual ownership chain.

What about employees and contractors?

Employees and contractors are treated differently, and the distinction matters.

For employees, IP created in the course of employment will often belong to the employer under general legal principles, but a well-drafted employment agreement should still spell this out clearly. That helps avoid disputes over side projects, inventions created partly outside work hours, or improvements made after the employee leaves.

For contractors, the default position is much riskier. Paying an invoice does not mean your startup owns the work. Unless the contract clearly assigns IP to your company, the contractor may retain ownership and only give you a limited implied right to use the work. That is not good enough for most investors or buyers.

How do AI tools complicate ownership?

AI products often rely on third party systems, and those systems may limit what you own or what you can commercialise.

For example, the licence for an underlying model, dataset, code library or API may restrict:

• commercial use
• modification or fine-tuning
• sublicensing to customers
• use in regulated industries
• ownership of outputs or derivative improvements
• training on customer data

This means your startup may own the application layer and customer relationships, while only having a limited licence to use key underlying components. That can still work commercially, but only if you understand the boundaries before you sign enterprise contracts or make promises about exclusivity.

Can AI-generated outputs be owned?

The answer depends on what the output is, how it was generated and what contracts or platform terms apply.

In practical terms, New Zealand businesses should not assume that every AI output will attract strong standalone IP rights. Where output value matters, founders usually rely on a mix of copyright in human-authored material, confidential information controls, database value, product design, customer contract terms and brand strength. If your product depends on exclusive ownership of generated content, that issue deserves specific legal review.

When This Issue Comes Up

IP ownership questions usually surface at the exact moment a startup wants to move faster. That is why they are often missed until the stakes are higher.

When you form the company

Many AI businesses start with a loose founder arrangement. One person writes code, another designs the workflow, and a third tests prompts with customers. If nobody documents ownership early, the company may not actually control the core product once it is incorporated through the Companies Office.

Before you sign founder documents, make sure they deal with pre-existing IP, future IP, confidentiality and what happens if someone leaves early.

When you hire developers, designers or ML specialists

The risk appears as soon as outside help is involved. A freelance developer may own code. A prompt engineer may own parts of the workflow documentation. A design contractor may own the product visuals and logo.

Before you launch online or pitch stockists, enterprise customers or channel partners, check that each contributor has signed terms covering IP assignment, moral rights consents where relevant, confidentiality, use of open source software, and restrictions on reusing your confidential materials elsewhere.

When you use third party data or APIs

Data sourcing is a common pressure point for AI startups. A team may scrape content, buy datasets, ingest customer records, or test with publicly available information without fully checking usage rights.

Ownership and privacy are different questions, but they often overlap. If personal information is involved, the Privacy Act 2020 may affect what you can collect, store, disclose and reuse for training or product improvement. Even if a dataset is technically accessible, that does not mean your startup has the contractual or privacy permissions needed for commercial use.

When you sign customer contracts

Customers often care less about your internal legal structure than about the outputs and improvements their money helps produce.

Enterprise deals commonly raise questions such as:

• Who owns customer input data?
• Who owns generated reports, recommendations or creative outputs?
• Can the startup reuse customer prompts or usage patterns to improve the model?
• Does customer feedback become part of the product IP?
• Will customer confidential information be used to train shared systems?

If your customer terms are silent or inconsistent, sales teams can accidentally promise ownership positions the business cannot legally support.

When you raise investment or prepare for due diligence

Investors usually want a clean story: the company owns the core IP, licences are understood, and there are no obvious disputes or missing assignments.

Due diligence often exposes issues that seemed harmless earlier, such as:

• a founder never assigned the MVP to the company
• a contractor built key features without a written contract
• an employee imported code from a former employer
• customer terms promise ownership of outputs that conflict with your platform terms
• the startup has invested in branding without checking trade mark availability

These problems are usually easier and cheaper to fix before the term sheet arrives.

Practical Steps And Common Mistakes

The safest approach is to build an ownership record from day one, then make your contracts match how the product actually works.

1. Create an IP map of the product

Founders often underestimate how much they are relying on. Write down the assets that matter commercially and where they came from.

Your internal IP map should cover:

• who created each core asset
• whether it was created before or after the company existed
• whether the asset is owned, licensed or shared
• what third party terms apply
• what customer data is used and for what purpose
• what protection method fits each asset, such as copyright, confidentiality, trade mark registration or contract rights

This exercise is especially useful before you invest in branding, register a business name or domain, or print packaging for any hardware-enabled AI product.

2. Use founder and contractor documents that actually assign IP

A vague clause about work being done for the business is usually not enough. The agreement should clearly state what IP is assigned, when assignment takes effect, and whether pre-existing materials are excluded or licensed in.

For AI startups, this should also deal with:

• training data preparation and labelling work
• prompt libraries and workflow rules
• model tuning and evaluation methods
• documentation, demos and sales materials
• rights to use or adapt pre-existing tools brought in by the contributor

If a founder or contractor wants to retain ownership of background IP, the business still needs a broad enough licence to use, modify and commercialise the product without disruption.

3. Review open source and platform terms early

The main risk is not just non-compliance. It is building a business model on assumptions the licence does not support.

Some open source licences may require certain disclosures or impose conditions if software is distributed in a particular way. Some hosted AI platforms may change terms, limit indemnities, or reserve rights around outputs and service data. Before you sign a major customer or reseller agreement, check whether your upstream terms allow the promises you want to make downstream.

4. Set clear customer positions on data, outputs and feedback

Your customer contract should reflect your product reality, not generic software wording copied from another business.

Depending on the model, the contract may need to explain:

• that the customer keeps ownership of its input data
• whether the customer owns final outputs or receives a licence to use them
• whether de-identified or aggregated usage data can be used for analytics and service improvement
• whether customer feedback can be incorporated into the platform
• what restrictions apply to customer use of outputs, especially where third party material may be involved

Clear drafting also supports compliance with the Fair Trading Act 1986. Marketing claims about ownership, originality, exclusivity or model training should match the legal reality.

5. Protect confidential know-how, not just formal IP rights

Some of the most valuable parts of an AI product are hard to register. Internal prompts, ranking logic, fine-tuning approaches, annotation guidelines, sales playbooks and deployment methods may be more realistically protected as confidential information.

That only works if the business treats them as confidential in practice. Use confidentiality clauses, limited access controls, sensible internal policies and clean offboarding steps when staff or contractors leave.

6. Think about trade marks early

AI founders often focus on code and ignore branding until later. That can be expensive if another business already has rights to the name.

Before you launch an online store, publish a product page or invest in design, check whether the brand is available and whether trade mark registration makes sense in New Zealand and any export markets. Owning the code does not help much if you need to rebrand after launch.

Common mistakes New Zealand AI startups make

Several patterns show up repeatedly in early stage businesses.

• Incorporating the company after the MVP is built, but never assigning the IP to the company.
• Using friends or offshore contractors without a written contract.
• Assuming payment equals ownership.
• Using datasets or scraped content without checking commercial rights or privacy issues.
• Promising customers exclusive or full ownership rights that conflict with upstream licences.
• Ignoring employee side projects and prior employer restrictions.
• Delaying trade mark checks until after money has been spent on branding.
• Keeping poor records of who contributed what, and when.

These are fixable, but they become harder once your startup has multiple products, outside investors, or long term enterprise customers.

How this fits with privacy, contracts and business structure

IP ownership is only one piece of the legal setup for an AI startup in New Zealand. It intersects with privacy disclosures, service terms, employment agreements, contractor agreements, shareholder arrangements and your business structure.

If you are looking to start a tech business in New Zealand, or scale an existing one, it helps to align those documents early. A company structure may be preferable for investment and asset ownership reasons, but the right setup depends on your situation. Accountants and tax advisers should be involved for tax questions, especially where IP is being transferred between people and entities.

FAQs

Does my New Zealand company own code a founder wrote before incorporation?

Not automatically. If the code was created before the company existed, the founder may still own it unless there is a written assignment to the company.

Do I need an IP clause if I hire a contractor to build part of my AI product?

Yes. Without a clear written assignment or licence, the contractor may retain ownership, even if your business paid in full.

Can my startup use customer data to train or improve the product?

Only if your contracts, privacy position and actual practices support that use. You need to consider the customer agreement, your privacy policy, and whether personal information is involved under the Privacy Act 2020.

Should I register a trade mark for my AI product name?

Often, yes. Trade mark registration can help protect the brand you are building, especially before you invest heavily in marketing, domains, packaging or customer acquisition.

Do AI-generated outputs belong to my startup or my customer?

There is no single default answer that suits every product. The position will depend on the technology, the relevant platform terms, the level of human input and, very often, what your customer contract says.

Key Takeaways

• AI product startups in New Zealand need to look beyond source code and identify all valuable IP, including data rights, prompts, workflows, branding and confidential know-how.
• Your company does not automatically own everything founders, employees or contractors create, especially work produced before incorporation or by contractors without written assignments.
• Third party models, APIs, datasets and open source software can limit what your startup owns and what you can promise customers.
• Customer contracts should clearly address ownership of inputs, outputs, feedback, improvement data and reuse rights.
• Privacy, trade mark registration, confidentiality protections and clean record keeping all support a stronger ownership position.
• Fixing ownership gaps early is usually much cheaper than fixing them during investment due diligence or a commercial dispute.

If your business is dealing with IP ownership AI product startups and wants help with founder IP assignments, contractor agreements, customer terms, trade mark strategy, you can reach us on 0800 002 184 or team@sprintlaw.co.nz for a free, no-obligations chat.