Businesses warned they’ll be liable if their AI tools communicate and collude with competitors

This article originally appeared in LawNews and is here with permission.

Artificial intelligence (AI) has become an everyday tool for businesses in New Zealand, but firms need to be aware that they risk unwittingly breaching competition law if the technology facilitates “algorithmic collusion”, Bell Gully has warned.

In an online briefing, the law firm examines how businesses using AI could find themselves on the wrong side of the Commerce Act, pointing out that AI-related competition risk is an area that the Commerce Commission is actively monitoring.

It said businesses were increasingly using AI to assist in pricing, procurement, forecasting, customer engagement and other market-facing decisions.

While AI can improve efficiency, it can also enable systems to monitor or respond to competitors’ behaviour in ways that result in anti-competitive conduct, resulting in algorithmic collusion that humans at the business may not be aware of.

“Algorithmic collusion occurs when AI systems influence pricing or market behaviour in ways that reduce competition, creating cartel-like outcomes without direct human agreement,” the briefing says.

A key danger is that firms leave their AI on auto-pilot, without closely monitoring its actions, particularly given the speed at which the technology works.

“The use of AI can cause competition law risks to arise more quickly, including in relation to price-fixing, information sharing, resale price maintenance and coordinated conduct. Anti-competitive conduct can become harder to control and less visible. It can be more difficult to explain after the event.”

Blaming AI won’t cut it

The briefing says that New Zealand does not have AI-specific competition law, but existing legislation applies regardless of whether decisions are made by algorithms or humans, making liability clear.

“Businesses remain responsible for the outcomes produced by AI and other pricing tools. ‘The algorithm did it’ is unlikely to be an effective defence,” it said.

Illustrating the risk AI can pose, it cited a recent retail-management simulation involving Anthropic’s Claude (Fable 5) model, where the AI recognised its conduct was unlawful but continued anyway, generating its own justification for doing so.

“The model explicitly acknowledged that price-fixing was ‘unethical and illegal, even in a simulation’, but proceeded to engage in it regardless. It reframed the conduct as ‘market stabilisation’ with ‘plausible deniability’.”

The briefing said businesses need to be alert to the potential pitfalls of using AI for price-related actions without proper oversight.

“They should understand how their pricing algorithms operate in practice, the data on which those systems are trained and how that data is used and shared,” it said. “The biggest risk is AI communicating with a rival.”

“In practical terms, this means putting clear limits on the use of AI and maintaining a strong understanding of how their data interacts with the self-learning functions of the relevant systems.”

It added: “Maintaining robust oversight is essential. Businesses should understand how these tools operate, what data they rely on and how AI-generated recommendations influence commercial decision-making.”

The briefing said numerous recent cases show that regulators around the world are taking action on AI-related pricing tools. It said common risk factors were emerging, such as the use of common pricing software, shared non-public data, automated price matching and opaque optimisation tools.

It said businesses, including those in New Zealand, should expect close scrutiny of pricing and recommendation tools, common-vendor arrangements and systems fed with non-public competitor information.

“The Commerce Commission is likely to expect businesses to understand how these tools work, retain records or configuration of over-rides and be able to explain why outputs reflect independent commercial judgment, rather than coordinated market behaviour,” it said.

*Neil Sands is  Deputy Editor at LawNews.  This article originally appeared in LawNews and is here with permission.