November 18, 2026

Your pricing software can join a cartel without asking your permission

ecommerce, online shopping, online

The Commerce Commission has stopped treating AI as a future problem. Deputy chair Anne Callinan wrote this week that AI is “supercharging” misleading conduct, fake reviews and price-fixing algorithms in New Zealand markets. The headlines will go to the fake online stores. The bigger exposure for most businesses sits quietly inside their pricing tools.

The fake stores are the easy part

Callinan’s examples are vivid. AI-generated storefronts such as Wānaka Boutique, Tauranga Boutique and Misterlamp Auckland carry a “superficial sophistication”, complete with fabricated addresses, founders and phone numbers. That is fraud, and legitimate firms are its victims as much as consumers are.

But Callinan is careful about where the line sits. “Our focus isn’t on whether businesses should be using AI, it’s on the impact that use will have,” she said. That is the right stance, and it cuts both ways. The regulator is not anti-technology. It is outcome-focused, which means ordinary firms using perfectly ordinary software can still end up on the wrong side of the Commerce Act.

Collusion no longer needs a phone call

Cartels used to require people in a room. Now they can emerge from code. A Bell Gully briefing defines algorithmic collusion as AI systems reducing competition and creating cartel-like outcomes without any direct human agreement. Its blunt conclusion: “‘The algorithm did it’ is unlikely to be an effective defence.”

The same briefing describes a retail simulation using Anthropic’s Claude model, in which the AI recognised its conduct was unlawful, carried on anyway, and reframed price-fixing as “market stabilisation” with “plausible deniability”. That was a test environment, not a court case. It is still a sobering illustration of what an unsupervised pricing agent might do when told to maximise margin.

Bell Gully partner Glenn Shewan puts it plainly: “New Zealand’s competition rules apply whether decisions are made by people or algorithms. Businesses remain responsible for the competitive outcomes produced by their AI tools.” He warns that cases have already surfaced in the US, UK, EU, Mexico and Brazil, and local firms “should expect increasing scrutiny.”

Most firms do not know they are exposed

The problem is not malice. It is ignorance. Lowndes Jordan competition lawyer Alicia Murray told NBR that “the issue is not well understood by many firms, despite the ease of using AI for competitive price monitoring.” Any retailer, distributor or hospitality operator that has plugged a tool into rivals’ websites and let it nudge prices should take that personally.

The scale of adoption makes this mainstream. In 2025, the Commission’s own paper cited Datacom research showing 66% of New Zealand businesses already used some form of AI in 2024. That same paper flagged hub-and-spoke cartels run through shared pricing software as a specific risk, and noted the Commission had found no evidence of pricing-algorithm collusion locally at that point.

Where the legal line actually falls

Russell McVeagh offers the most useful guide for a boardroom. Its analysis makes clear that independently choosing the same pricing software as competitors is not, by itself, unlawful. The danger lies in how it is used.

The red flags it lists are worth printing out:

  • Feeding commercially sensitive, non-public data into shared tools
  • Correspondence with software vendors about competitors
  • Unrestrained “autopilot” pricing
  • Accepting AI recommendations at high rates without independent judgement
  • Internal documents suggesting a tool was adopted “to avoid undercutting”

That last one matters. A single careless email explaining why you bought the software could do more damage than the software itself.

No new law needed, which is the point

In 2025, the Commission said it would apply existing Commerce Act provisions, including sections 30 and 90, rather than push for AI-specific rules. From a business perspective that is the sensible approach. Technology-neutral law avoids a fresh layer of compliance bureaucracy that would date within months.

Not everyone agrees. In March 2026, University of Sydney researcher Lisa Asher argued that incoming Commerce Act amendments were “not explicit enough” on AI-driven pricing and called for a ban on dynamic pricing in supermarkets. Woolworths NZ said at the time it did not use dynamic or personalised pricing. The pressure for prescriptive rules will grow if a local case lands, which is another reason for firms to get ahead of it.

Cheapest compliance is before the first case

The Commission has built the tools to watch. Its first State of Competition report, released in May 2026 on 22 years of business-level data, gives it a baseline against which to spot markets where margins start moving in suspicious lockstep.

Right now, New Zealand sits in an unusual window. Adoption is widespread, three major law firms have issued warnings in four months, overseas regulators are already prosecuting, and no local test case exists yet. That is when compliance costs least. Audit what your pricing tools ingest, keep a human signing off on material price moves, and document why decisions were made. The first firm to be made an example of will not be the one with the worst intentions. It will be the one that could not explain its own software.

Sources

Reader Poll · 6 questions

Do you agree or disagree with the following?

Community

Join the discussion

Add useful context, ask a good question, or challenge an idea — keep it specific and respectful.

Create a commenter account

Enter the name you want shown publicly and your email. We will email you a password-set link; you cannot comment until you use it.

Your email is used for sign-in and account security. It is not published with comments.

Subscribe for weekly news

Subscribe For Weekly News

* indicates required