Skip to main content
AIGovernance Risk and Compliance

Accountability doesn’t shift to the algorithm: what the Commerce Commission’s AI guidance means for New Zealand boards

By November 21, 2026No Comments

On 18 November 2026 the Commerce Commission published guidance on how the Fair Trading Act 1986 and the Commerce Act 1986 apply when businesses use artificial intelligence (AI). It is the clearest statement yet from a New Zealand regulator that AI governance belongs in the boardroom.

The Commission is careful to say it does not regulate AI tools or systems, and that it supports their use. What it regulates is conduct. If an AI system makes a misleading claim, sets a price in a way that harms competition, or pressures a customer into a purchase, the business that deployed it is responsible. As Deputy Chair Anne Callinan put it, “Using AI does not shift that accountability and responsibility.”

For mid-market organisations that have adopted AI quickly and informally, that is a significant shift in how the risk should be framed.

Existing law, new ways to breach it

None of this is new law. The guidance applies long-standing obligations to new technology, and that is precisely why it matters. There is no transition period and no threshold to wait for.

Under the Fair Trading Act, the Commission identifies a range of AI-specific risks. AI-generated advertising, website copy and chatbot responses can be misleading or deceptive. Claims produced by AI still need reasonable grounds, and the guidance says plainly, “Don’t assume AI-generated content is accurate.” Synthetic images, video or audio used as endorsements, AI-generated reviews, and personalised pricing that creates a false impression are all in scope. So is “AI washing”: describing a product or service as “AI-powered” when the claim overstates what it does.

Under the Commerce Act, the focus shifts to pricing and competition. Algorithms that coordinate prices, shared pricing platforms that pass sensitive information between competitors, and autonomous systems that drift towards coordinated outcomes can all amount to cartel conduct. The Commission is explicit that the cartel provisions apply whether a decision is made by a person, by software or by AI.

The penalties are not trivial. Under the Fair Trading Act, businesses face up to $600,000 per breach. Under the Commerce Act, penalties for a business can reach the greater of $10 million, three times the commercial gain, or 10 percent of turnover, and individuals involved in cartel conduct face fines and up to seven years’ imprisonment.

Where mid-market organisations are exposed

Large enterprises tend to have legal review built into marketing and pricing. Mid-market organisations often do not, and AI has removed much of the friction that used to slow content and decisions down.

The exposure usually sits in three places. The first is customer-facing content, where AI now drafts product descriptions, campaigns and social posts at a volume no one reviews line by line. The second is customer interaction, where chatbots and agents answer questions about price, availability and terms without a person checking what was said. The third is pricing, where third-party tools adjust prices dynamically using data the business has never examined.

In each case the tool was usually adopted to save time, by a team that reasonably saw it as an operational decision. The Commission’s guidance reframes it as a compliance decision.

What boards should ask for

The Commission says boards should have visibility of AI use and build consumer protection and competition law into AI governance and risk frameworks. In practice, we recommend Directors request four things from management.

A register of AI in use. Every tool that generates customer-facing content, interacts with customers or influences price should be listed, with an owner and a description of what it does. If management cannot produce this, the board cannot oversee it.

A review step for AI-generated claims. Marketing and product content produced with AI should pass through the same substantiation check as any other claim, with records kept. The guidance specifically calls for records that support the claims a business makes.

Testing of customer-facing tools. Chatbots and agents should be tested regularly against realistic customer questions, including questions about price, refunds and guarantees, with results reported to a named executive.

Scrutiny of pricing tools and suppliers. For any algorithmic pricing tool, management should be able to explain what data it uses, whether competitor data feeds its outputs, and how a person oversees its decisions. Supplier contracts should answer the same questions.

Governance that enables adoption

None of this argues for slowing AI adoption. The Commission is clear that it focuses on harmful conduct, not on the technology. Organisations that can show how they govern AI will move faster with more confidence, because they will know where the boundaries sit.

The practical next step is to add AI-related consumer and competition risk to the risk register as its own item, assign an executive owner, and bring the register and controls above to the board within the next quarter.

Unisphere helps boards and leadership teams put proportionate AI governance in place without stalling progress. Our AI readiness assessment identifies where AI is already operating across the business and what oversight each use requires. To discuss how the Commerce Commission’s guidance applies to your organisation, talk to our AI and digital transformation practice or visit unisphere.co.nz.

Sources: Commerce Commission media release, 18 November 2026; Commerce Commission, “AI guidance for businesses” (AI and the Fair Trading Act; AI and the Commerce Act; Algorithmic Collusion and the Commerce Act), November 2026.