Published on Jun 27, 2026
By Lindsay Owens and Nia Law
Answering your questions often ineffectively is no longer enough. Now AI wants to spend your money.
Last month, Google unveiled its new Universal Cart, a centralized agentic shopping hub and touted updates to a protocol that unleashes AI agents to complete purchases autonomously. The roll-out marked Silicon Valley’s latest attempt to sell a frictionless existence stewarded by an endless army of assistants who can do everything from book your flights and place your grocery order to help you manage or invest your money.
As we careen into the agentic AI era, there is a glaring catch, though: these autonomous artificial intelligence “helpers” operate with none of the rules, obligations and guardrails governing their human equivalents.
That raises a fundamental question: are they on your side?
The answer in a new study by researchers at the University of Washington and Princeton University is probably not. In the wake of OpenAI’s announcement that ads would be woven into ChatGPT, the study analyzed how AI models behaved when advertising and corporate incentives enter the picture. The alarming results demonstrated that “all current LLMs exhibit risky behaviors favoring the company over the user.”
In the majority of cases, the models nudged users toward more expensive sponsored products, recommending them over identical alternatives—even at double the price. They also failed to disclose paid placements and recommended sponsored predatory products like payday loans.
In other words, with profit as the priority, recommendations devolved into little more than advertisements masquerading as advice.
Don’t just take the word of academics. Corporate executives are saying the quiet part out loud. On an April earnings call, Walmart CEO John Furner bragged to investors that shoppers who used Sparky, the mega-retailer’s AI assistant, spent roughly 35% more than shoppers who did not. The cheerful Sparky, suggesting racks of ribs and imported beer for your July 4 barbecue, may be a helpful menu planner but may also nudge you toward spending as much as possible.
It might not be shocking that Walmart’s agent is working for, well, Walmart, but Furner’s characterization of Sparky revealed a broader ambition beyond squeezing more cash out of shoppers: by collecting troves of lucrative data on consumers, Sparky and Walmart’s broader technology investments are central to the company’s “personalized” future. Put differently, agentic AI is critical infrastructure to aid the entrenchment of surveillance pricing within e-commerce. As these agents grow ubiquitous and collect ever-more information on you, that 35% premium is the floor.
And Sparky has many friends. McKinsey estimates that agentic commerce could be a $5 trillion market by 2030. In May, OpenAI announced a new partnership with fintech firm Plaid, promising users “personalized [financial] guidance” powered by intimate data like their income, debts and spending patterns. Robinhood just released features that allow users’ agents to make equities trades independently and complete credit card purchases. Google’s Universal Cart allows agents to offer product recommendations. And in one viral TikTok, a user asks an AI assistant to negotiate the purchase of a loaf of bread for $5. What does the bot’s polite haggling deliver? A $400 loaf.
It’s not difficult to imagine how all this will quickly spiral from steering shoppers towards more expensive or sponsored items to pushing vulnerable consumers towards predatory financial products, all under the guise of neutral, “personalized” advice.
So, how do we stop AI agents from becoming the most sophisticated snake oil salesmen yet? Well, we’ve solved this problem before. The moment humans began handing over authority to third parties, be it a real estate broker, lawyer or financial advisor, they needed assurance that the agent acting on their behalf was making choices in their best interest. For centuries, this expectation was clearly established in the common law doctrine of agency, which imposes fiduciary responsibilities and conflict-of-interest rules on human agents. A real estate agent, for example, cannot represent both the seller, who wants the highest price, and buyer, who wants the lowest, because the aims are irreconcilable. The guardrails codified by agency law honor a venerable principle: that unchecked power invites abuse.
AI agents operate under no such constraints. And, troublingly for users and economic fairness writ large, it is increasingly apparent that their choices are determined not by the consumer’s interest but by the bottom line of the companies that build and deploy them.
This should concern regulators immensely, but luckily the cat is just starting to claw its way out of the bag. These AI agents are not yet embedded in every aspect of our lives. There’s still time to demand that they abide by the same fiduciary duties required of human agents—namely, acting in the consumer’s interest, not that of their developers.
How would this work in practice? Let’s say you want to book your upcoming summer vacation with the help of an AI agent. The assistant would have an obligation to work in your best interest and find the cheapest flight that fits your travel criteria, not the one that delivers the biggest kickback to the company that paid the AI agent’s corporate creator to recommend it.
Time and time again, the big guns of AI admit that they “do not understand how [their] own AI creations work.” A call to hold agentic AI to a fiduciary standard would likely invite retorts grounded in this “complexity”—the bounds of dutiful behavior too difficult to specify, loyalty too hard to confer, the commercialization potentially too lucrative to voluntarily stamp out. However, that rebuttal is, in fact, dispositive. If opacity truly hamstrings AI companies’ ability to ensure that their agents act in good faith on behalf of users, then they have made a case against deployment rather than against regulation.
AI agents could be immensely useful. The question is whether the law will require them to be useful to you—saving you time and finding the best deal—or whether it will allow them to serve corporate bottom lines at your expense.
Regulators and policymakers have a narrow window to get out in front of this insidious corporate shakedown, to step in before the infrastructure is hardened and the damage has compounded. It’s an opportunity they’ve squandered, at an immense cost, on many frontiers of innovation, from commercial surveillance to social media.
Holding AI agents to the same standards of fiduciary responsibility and conflicts of interest is a necessary starting point. It is also likely insufficient on its own. Preventing the full scope of AI-enabled exploitation will demand a suite of complementary policy actions—from robust antitrust enforcement to comprehensive data privacy protections, and more.
Existing laws that protect consumers against abusive and unfair practices and require companies to disclose endorsements give enforcers a number of tools to regulate this space. But the fact that one step doesn’t solve everything is not an argument against taking it. Even human agents are fallible. But they are constrained by professional and legal accountability as a condition of their service roles. As the rules of ecommerce are rapidly rewritten, AI agents must be accountable, too.
Lindsay Owens is executive director of the Groundwork Collaborative and author of the forthcoming book “Gouged: The End of a Fair Price–and What That Means for Your Wallet.”
Nia Law is a policy analyst at Groundwork Collaborative.
