ACCC warns of personalized pricing and colluding bots: Australia's regulator is lagging behind shopping agents

By The Nightly (AAP) · October 16, 2026.

The Australian Competition and Consumer Commission (ACCC) raised concerns on Friday before a parliamentary committee about the use of AI in pricing. Three fronts are of concern: personalized pricing, "agentic shopping" (purchases delegated to bots) and AI-powered scams.

According to David Abkiewicz, the ACCC's general manager of digital markets, the real harm appears when personalization is "extremely granular": it can go as far as establishing each customer's maximum willingness to pay, or rely on data that reveals vulnerability. Charging different prices to different groups is not against Australian consumer law, he acknowledged, though he noted that other jurisdictions, such as the US, are beginning to restrict "surveillance pricing" (prices based on monitoring of personal data). By the regulator's own account, there is no simple avenue to act against companies that exploit vulnerable customers with this practice.

The second warning is more novel. As more consumers use AI agents to shop on their behalf, consumer protection bodies fear those agents could carry built-in brand preferences or even "collude" to steer shoppers away from the best deals. Abkiewicz spoke of continuing to monitor that agents "work in the interests of consumers themselves." This is a concern voiced by the regulator, not a proven finding.

Third front: scams. Jayde Richmond, of the National Anti-Scam Centre, said AI makes it possible to operate at scale and to automate and personalize contact. It is not yet being reported in significant numbers, but they are already seeing voice cloning, deepfake videos and possible manipulation of search results. Abkiewicz added that, without intervention, the digital economy tends toward dominant players that are hard to dislodge, and that we are at a "very early" stage.

The most visible action so far: the ACCC sued Microsoft for allegedly misleading 2.7 million Australians about pricing when it integrated Copilot into its flagship consumer product. This is an allegation by the regulator, pending a court ruling.

Our take: the pattern is that the law was written for a market where the price was a label, not an individual conversation. Therein lies the thesis: the risk of AI in commerce is not only that prices go up, but that they stop being comparable. If every customer sees a different price and their shopping agent is opaque, the benchmark that makes competition work disappears.

The warning about bots that "conspire" is, for now, hypothetical and should be treated as such. But it points to a real, plumbing-level problem: whoever controls the agent that decides what to buy controls a bottleneck, just as happens with agentic payment standards. In general, the sensible response is not to ban but to demand transparency: that the agent declare on whose behalf it acts, with what incentives, and which alternatives it discarded.

In the short term there will be friction: vulnerable consumers exposed, regulators reacting late. In the long term, an agent truly loyal to the user could be one of the best tools for regaining bargaining power against companies. That is the version of abundance worth defending, and it depends on regulation requiring that the agent work for you and not for the seller.