Navigating the intersection of AI and Intellectual Property

The rise of generative artificial intelligence (Gen AI) has had a major impact on a range of industries, with clear benefits and efficiencies to be gained, and challenges to be overcome.  Many software companies are moving quickly to embed Gen AI into their Intellectual Property (IP) stable, across product development, engineering workflows, and customer-facing features; but that pace brings pressure — and risk.

Our clients are increasingly asking:

• Who owns the intellectual property (IP)?
• Are we exposed to liability?
• Can we protect the value we're creating?

These are not hypothetical questions. Globally, we’re tracking a wave of regulatory reform, litigation, and investor scrutiny that’s reshaping how IP, data, and liability are treated in AI-heavy businesses. We have seen IP ownership and model usage rights become gating issues in M&A transactions. In audits, questions are being asked around developer practices, tool governance, and model transparency. Certainty around what assets are capable of IP protection, making sure that you own those assets, and ensuring you are protected from the risks and liabilities associated with AI use, are increasingly becoming critical for the value of your business. 

Safe, scalable Gen AI adoption can be achieved through practical controls, enterprise-grade environments, and clear governance. For software companies in particular, that includes helping teams get confident about the IP and compliance implications of AI adoption.

Whether software code generated through the use of AI qualifies for IP protection is not yet clear as it remains unsettled law, and the protection available differs from country to country. Internationally, software is primarily protected by copyright in the underlying software code - a protection that applies automatically when a person creates an original piece of code. Additionally, in some countries, businesses can protect their code through patents.

New Zealand

New Zealand copyright law allows AI-generated works to qualify for copyright protection, with the default protection being granted to the ‘author’ - being the person by whom “the arrangements necessary for the creation of the work are undertaken”. In practice, this means that businesses will need to show active involvement by a human in generating the work to establish copyright (through, for example, conscious input as prompts).

International position

The position internationally is less clear, and can lead to issues for businesses trying to establish or enforce their rights. For example, like New Zealand, the United Kingdom allows AI-generated works to qualify for copyright protection, but it is expected that active human involvement in generating the work would be required to establish copyright.  

On the other hand, both Australia and the United States require a human author for copyright to arise. Developments in the United States indicate that copyright does not extend to purely AI-generated material. Copyright can protect some human generated work that also includes AI-generated material, but whether human contributions to AI-generated outputs are sufficient to constitute authorship will be considered on a case-by-case basis - which isn’t helpful for businesses looking for certainty. The Australian courts have not yet tested the extent to which work created by a human that incorporates AI-generated content could or would be protected.

Key takeaway

New Zealand businesses may find that the rights they enjoy in New Zealand do not exist elsewhere. Where aiming to secure international IP protection is a priority, it is crucial to provide clear guidance to developers around AI use in developing novel code and record keeping. Ensuring a significant level of human input and authorship is retained while utilising Gen AI tools and documenting human contributions in both prompt authorship and subsequent refinement of AI-generated output will strengthen claims for copyright protection internationally. 

Who owns IP rights in AI-generated content?

Generally, and across some of the more ubiquitous AI tools being used, the content prompt author will own the IP in what you input into the tool, and they will own the outputs as well. That being said, there is no uniform standard or approach across tools or legislatively. 

Key takeaway

Contractual arrangements with AI tool providers should be checked to ensure IP ownership provisions are appropriate, both in relation to what you input into the tool, and the outputs it generates.

Infringement of third party rights

The risk of infringement of IP and open source rights will depend on what content has been fed into the tool, what authorisations were given in relation to this content, and whether the AI-generated content is sufficiently different to the original content to avoid infringement. There are a number of cases before the courts internationally currently considering these issues. 

Key takeaway

A common sense approach should be taken when reviewing AI-generated outputs - if the output is a clear duplication of bespoke third party content - don’t use it. Examples include outputs with domain specific information or code that still has comments attached to it.

In addition, to help mitigate these risks, be mindful of the tools you use and the information available on how those tools were built. Check your contracts for the AI tools you are using for the protection provided (or lack thereof) to you in a third party infringement claim. Many tools will provide an IP infringement indemnity, under which the tool will manage infringement claims made against you for use of outputs and cover your associated costs.

Using AI tools may make your IP and data input into these tools vulnerable to being shared with or leveraged by third parties. 

Keeping IP assets confidential provides an additional layer of protection. For example, software code can be protected as both a copyright work and as confidential information - if the code is confidential, the risk of someone copying the code is (obviously) reduced. If the code is scraped by or fed into an AI tool, there is a risk that the code could be shared with others.

Another way to mitigate this risk is to implement procedures to ensure that AI tools used by the business have appropriate security controls; and that the contractual terms and privacy policies governing use of these tools provide appropriate protection for data inputs and outputs. It is safer to use an enterprise version of an AI tool that does not share your inputs and outputs outside your organisation, and you should work with authorised resellers to ensure that you get the comfort you need for your valuable IP assets. 

Your own internal policies and team training are also useful to mitigate the risk of key assets being unwittingly shared.

Future outlook

The rapid development of Gen AI is spurring governments across the world into action. New Zealand has indicated its intention to take a “light touch, proportionate and risk-based” approach to regulating AI, however, as this area is constantly, and rapidly evolving, there is a real risk that the legislation, when implemented, is no longer fit for purpose. Internationally, countries are enacting specific AI legislation to address the novel challenges of Gen AI use and regulation, and while often well considered, they face the same challenge of regulating exponential change. 

With Gen AI-related disputes increasingly appearing before the courts, frequently engaging with legal professionals and AI experts is important to ensure you remain educated and protected in this evolving legal landscape. 

There are great benefits gained by using these tools, but there are also some risk points and therefore caution / being deliberate as to the tools you use and what you use them for is key. Generally, enterprise versions of tools are safer to use and provide better protection.

Businesses should consider, and have policies around:

• What AI generated content is used - for example, if the generated content clearly appears to have been lifted from somewhere else. 

• How AI generated content is used - if the generated content is intended to be used for something critical to the business or something that will be difficult to change in the future, you may wish to exercise greater caution in using that content.

• What content is input into AI tools - consider whether you want your business’ most valuable / unique content to be accessible to assist others in their development, and whether parameters are included in your AI policy around the scale of inputted code.

What contractual arrangements are in place - check your contracts with AI tool providers to ensure appropriate IP ownership, liability and security arrangements are in place.

Businesses that are using AI in the development of proprietary systems should keep up to date with developments in this area of law, and take steps to protect themselves whilst these issues and queries are being considered. We see clients that have well-developed AI and IP policies being able to navigate through sale processes with significantly less friction than those without (including reduced risk associated with IT due diligence processes). 

The integrated PwC AI and PwC Legal teams have been working with clients to deliver cutting-edge Gen AI solutions. PwC’s broad range of experts including lawyers, consultants, engineers, developers, data scientists, and design specialists, together with our alliances with leading tech companies and academic institutions, are perfectly placed to assist you in your Gen AI journey. From establishing effective AI governance to implementing bespoke AI capabilities, we are excited to help unlock the power and capture the value of Gen AI.  

For a broader view on the considerations for businesses around the use and governance of AI, see our report Artificial intelligence: What directors need to know