The gap between adoption and law is widening
New Zealand businesses are building workflows, generating marketing copy, writing code, and producing design assets with generative AI tools. Two-thirds of larger firms were already using some form of AI in 2024, up from 48% the year before. The government’s own strategy document projects generative AI could add $76 billion to the economy by 2038.
But here is the problem. The Copyright Act 1994 was drafted over 30 years ago and makes no express provision for generative AI. Whether AI outputs qualify as original works, whether training models on copyrighted material is lawful, and who carries liability when outputs infringe third-party rights are all unresolved questions in New Zealand law.
Commerce and Consumer Affairs Minister Cameron Brewer has acknowledged the issue. His public commitment amounts to this: “Cabinet has invited me to report back by 31 March 2027 on a possible copyright framework for generative AI in New Zealand.” That is a timeline for advice, not legislation. Actual law is years away.
You might not own what your AI produces
New Zealand’s Copyright Act is relatively anomalous internationally because it does recognise copyright in “computer-generated works.” The US, Australia, and EU all require human authorship. The UK is now proposing to remove protection for works created solely by AI while retaining it for AI-assisted works.
New Zealand has not taken a position. That means a business generating content with AI today has no certainty about whether it owns that content. The test under the existing Act hinges on the degree of human skill, labour, and judgement applied, a standard that is genuinely unclear for most commercial AI use cases.
Dr Joshua Yuvaraj, a law researcher at the University of Auckland, argued in February that “copyright law must reflect the lopsided division of intellectual labour in AI-assisted creation” and proposed that AI-generated works should enter the public domain sooner. He called it “an urgent moment” for reshaping the law. The government’s March 2027 report-back date does not suggest it shares that urgency.
Your vendor’s terms may already have claimed it
Even where NZ law might grant copyright in AI outputs, most generative AI providers impose terms of use that include assignment provisions, licences allowing the provider to use outputs for training, or restrictions on commercial use. A firm using ChatGPT or Midjourney for client-facing work may be operating under terms that effectively give the platform rights over the output.
Most businesses have not reviewed those terms. That is a commercial risk hiding in plain sight.
The training data litigation bomb
Global litigation against AI developers over whether training on copyrighted material constitutes infringement remains unresolved. In December 2025, Professor Graeme Austin of Victoria University’s law faculty wrote that “authors’ livelihoods are at risk” as generative AI platforms scrape massive amounts of protected material. He warned that if fair dealing exemptions were applied to AI training, “copyright owners would basically become unwilling donors of AI firms’ seed capital”.
Austin’s preferred solution is licensing markets, analogous to how bars pay to play background music. That is a market-based answer that should appeal to any centre-right government. But it requires a legal framework to support it, and that framework does not exist.
SMEs are sitting out, and uncertainty is part of the reason
The adoption data reveals a bifurcated market. While large firms race ahead, 68% of NZ SMEs have no plans to evaluate or invest in AI, compared to just 38% of Australian SMEs. Resource constraints explain part of that gap. But legal uncertainty is a genuine deterrent for small business owners who cannot afford to absorb a copyright dispute.
MBIE’s own voluntary responsible AI guidance, published in July 2025, explicitly flagged copyright compliance as a live obligation and called on businesses to conduct licensing assessments. The Employers and Manufacturers Association has warned its members that “blaming AI is not a legal defence if something goes wrong.”
Voluntary guidance and industry warnings are not substitutes for law. They are symptoms of its absence.
What businesses should do before the government catches up
The recent copyright changes announced by Brewer, extending sound recording copyright to 70 years, strengthening digital lock protections, creating a framework for blocking piracy sites, are real but largely unrelated to AI. They were driven by FTA obligations with the UK and EU, not by the generative AI challenge.
In the absence of clear law, businesses deploying AI commercially should review vendor terms of use before relying on outputs, document human creative input to strengthen any copyright claim under the existing Act, assess exposure to the EU AI Act if operating in European markets, and watch the outcome of US and UK training data litigation closely.
The government has projected a $76 billion prize from AI adoption. It has not yet built the legal infrastructure to let businesses claim it safely. Every month that passes without a framework is another month of commercial risk that firms are carrying alone.
Sources
- Newsroom: Do you own your AI-generated content? (2026-05-27)
- NZ Herald: Split Enz to benefit from Government copyright changes, Cameron Brewer to look at copyright in the AI era
- University of Auckland: Can AI be truly creative and what does this mean for copyright? (2026-02-25)
- Scoop: Creative and cultural sector gets further copyright support (2026-06)
- Employers and Manufacturers Association: The legal and regulatory framework of AI