With more companies and users creating content and images using generative AI (GAI), it is unclear however who owns the resulting copyright in these works.

Under the Copyright Act 1994 (the Act), computer-generated content (which includes content produced by GAI) is protected by copyright in New Zealand. This protected content includes materials such as written reports, design drawings, and computer code.

The Act states that the owner of the copyright is the author of the work, being the person who ‘made the arrangements necessary’ for the creation of the work. In the context of a GAI, it is unclear whether this is the creator of the AI programme or the user of the AI programme.

Concerns for businesses using AI-generated content

The practical implications of a business not having ownership of utilised content from a GAI can include:

• An inability to monetise business goods which incorporate this content;
• Issues with enforcing Intellectual Property (IP) rights in business goods which incorporate this content; and
• Issues with customers being able to own the IP in business goods which incorporate this content.

For businesses considering or using GAI, care should be taken where the business wishes to claim ownership of the output from the GAI. Where a business develops and uses its own AI, the business is clearly the owner of any works produced by the AI. However, the ownership of the output from public GAI is unclear.

Recommendations

Businesses using public GAI should have practices in place, which support their claim of ownership should such issues arise.

• Review the GAI’s terms of use. Some GAIs may state in their terms of use who owns the output from the GAI. For example, OpenAI (the developer of ChatGPT) assigns ownership of output from its GAIs to its users, as long as its terms of use are followed.
• Businesses should ensure there are clear employee policies in place setting out where generative AI may be used for business projects.
• When GAI is being used, users should keep a record of all input they enter in the software. This record may assist in proving the business was the one who ‘made the arrangements necessary’ for the creation of the work and is therefore the owner of the work.
• Contractor agreements with third parties should specify whether GAI may be used and the extent of such use. Where contractor agreements allow the use of GAI, the outputs generated from this use should be specified. The contractor agreements should also include provisions that, should any infringement issues arise from use of the GAI content, the contractor agrees to defend and indemnify the business.

As an example, software developers may use GAI to assist in writing code for new software. Therefore, based on the above, businesses using internal or external resources to develop software should have clear policies/contract terms around the extent to which AI can be used to write code. These practices will help protect against any ownership issues when the business later sells or licenses the software.

The Act is currently under review and part of the review process focuses on the ownership of computer-generated copyright works. Overseas, the position is increasingly being adopted that AI-generated content cannot be protected by copyright. The United States Copyright Office recently released guidance concerning the copyright status of generative AI works, which states these ‘works are generally not protectable by copyright as copyrightable works have a human authorship requirement’. It is not clear the extent to which this decision may inform changes to New Zealand’s copyright law, however this will clearly be a space to watch.

Infringement risks from use of AI-generated content

GAI’s use of training data sets can have infringement risks if the output from the GAI uses and/or collects data in a manner that infringes others’ rights. For example, Getty Images is currently suing Stability AI, the creator of an AI called Stable Diffusion, for infringing its copyright by using Getty photos to train its AI. While this indicates AI developers may be liable where GAI is using copyrighted material without consent as part of their training, this may extend to businesses utilising the output of these AI.

Recommendations

To minimise the risk of infringement issues, businesses utilising GAI should be mindful of the training data sets used by the AI.

• Businesses should only use GAI that have well documented training data sets and have obtained the relevant consents and/or licences from third party contributors to the training data sets.
• Where necessary, businesses may have to obtain licences and consents from the contributors to the training data sets before using the GAI.
• To ensure these processes are followed, they should be included in the employee policies referred to above.

If you or your business is unsure of the infringement risks of using GAI or how to assess these risks, get in touch with our experts for professional IP advice before using the AI.