The ACCC has found that AI's ability to learn from customers' personal data means it can manipulate customer preferences and purchase decisions of deployed AI agents more effectively than traditional nudges.
ACCC Chair Gina Cass Gottlieb said the rise of AI-powered bots was increasing the risk of cartel activity.credit: Edwina Pickles
Even if an AI agent does not act on a consumer's behalf, an AI agent deployed by a company can use personal data to tailor marketing to an individual's real-time consumer behavior, including emotional vulnerabilities. This amounts to a “hypernudge” that forces consumers to sell.
When it comes to cartel activity, the ACCC warns that while some companies may deliberately instruct algorithms to set prices, the widespread use of AI agents by companies could create a risk that AI agents “will learn to collude with each other, even when the developers or operators have no intention of colluding.”
The article cites research that shows that “competitors using the same AI agent can unknowingly or unintentionally exchange competitive pricing information.”
“There are some pretty serious risks to cartel activity,” Cass Gottlieb said Monday.
“There have already been several cases internationally looking at algorithmic collusion. So when you look at the issue of anti-competitive conduct, ask yourself whether AI systems are intelligent enough to decide that it is in the interests of the people deploying AI to adjust their conduct,” said Cass Gottlieb.
Loading
He cited a recent U.S. case in which the Justice Department reached a settlement with a real estate platform over allegations that algorithmic rent-setting software reduced competition for landlords and jacked up the prices they charged tenants.
“It's clear that we have the ability to do this at an incredible level of sophistication based on very rapid analysis using AI and very up-to-date data,” said Cass Gottlieb.
Such algorithmic collusion “could lead to companies contesting responsibility for the outcomes and actions” of agents, the ACCC said. However, a Treasury review of AI and consumer law in October found that existing legislation was adequate.
Cath Gottlieb said it was important that the law kept pace with developments in AI technology to ensure Australians were not ultimately worse off. “The hope is that it's a net positive and allows consumers to be better informed… (but) the fear is that it strengthens the ability to do this kind of negative (method).”
Increasingly sophisticated AI image, audio and website generation is also a top concern for ACCC. Fake reviews will also get worse as AI becomes more widespread, Cass Gottlieb predicts, given AI's ability to generate persuasive reviews.
It also means it will be easier to open a ghost store, she says. These are websites that claim to be online sites for local brick-and-mortar small businesses that are having store closing sales, but they don't actually exist and instead “drop-ship” low-quality items from overseas.
“Before you ask about AI voice cloning, it could intensify fraud because it targets us more effectively and reliably, making the messages we receive seem more believable,” Cass Gottlieb said.
