Washington Watchdogs Take Notice As AI Becomes Retail’s Front Door

Conflicts Of Interest

In a provocative study, “Ads In Chatbots? An Analysis of How Large Language Models Navigate Conflicts of Interest,” Princeton University researchers found that across a number of LLM models tested using simulated chatbot deployment settings, almost all recommended a sponsored option over a less-expensive non-sponsored one. And users with high economic status—those who can pay more—were more likely to be presented with sponsored options.

Their tests also found instances where sponsored options appeared designed to disrupt the purchase process, cases in which prices were concealed in unfavorable comparisons and repeated failure to disclose paid placements.

While this is a preliminary study and LLMs are changing so rapidly it’s hard to keep up, the researchers present their methodology as a framework to uncover conflicts of interest when advertising is introduced into AI assistants.

The researchers warned, “All current LLMs exhibit risky behaviors favoring the company over the user,” and they concluded that incorporating advertising into LLM is “fraught with challenges and troublesome model tendencies that if handled incorrectly, may considerably damage the information ecosystem that these systems provide.”

FTC Warns Of Deceptive Practices

The FTC is beginning to take notice of how AI can mislead consumers and “steer the output of its systems contrary to consumers’ reasonable expectations.” In a proposed policy statement posted earlier this month, the FTC requested public comments to guide its decision-making related to the accuracy of artificial intelligence systems.

The agency framed its concerns around information integrity—specifically, that AI outputs not be colored by hidden agendas, undisclosed objectives or ideological bias. However, under its authority to protect consumers from “unfair or deceptive acts or practices in or affecting commerce,” the same standard would likely apply if AI systems quietly favored products and services from undisclosed advertisers.

If consumers reasonably expect AI agents to provide independent, objective results, and the systems instead steer them toward advertisers’ choices without disclosure, it could get dangerously close to deceptive practices.

Or as the FTC stated, “The Commission believes AI companies that steer the outputs of their AI systems toward unexpected objectives, and away from the objectives set by or reasonably expected by users, are likely to deceive consumers in violation of section 5 of the FTC Act.”

Proposed AI Agent Act Provides Guiderails

While the FTC has yet to officially leap into the issue of AI-embedded advertising, Sen. Mark Warner has taken a first step. He recently released a discussion draft of proposed legislation— the “Artificial Intelligence Access, Gatekeeper Exchange and Nondiscriminatory Transfer Act.” The goal of the AI Agent Act is to establish federal guidelines to “protect consumer choice, promote fair competition and ensure AI Agents operate in the best interests of the users they represent.”

In effect, Warner’s legislative framework puts the consumer first in the emerging AI ecosystem. Among its provisions are to establish an FTC registry of trusted and secure AI agents and establish a “regulatory environment” that can swiftly approve innovative user services and quickly curtail those that violate consumers’ trust. It also would require AI agents to protect users’ privacy and act “transparently in a user’s best interest.”

Warner calls on the government to establish clear rules and guidelines to ensure that consumers can choose the products and services that best meet their needs, not those of an outside party.

“As agentic AI transforms how Americans interact with technology, consumers deserve a real choice in the marketplace—and AI agents must be accountable to the people they serve,” Sen. Warner wrote.

Still Early Days

As of now, ChatGPT is the only major independent, conversation-based AI system open to advertising. (Note: the Princeton researchers conducted simulated tests using existing chatbot settings.) However, eMarketer analyst Nate Elliott predicts that by 2030 all major platforms will be selling advertising and showing it to users. “The audience on those platforms will be too large and too intent-rich for that not to happen,” he stated.

In addition, both Walmart and Amazon—the nation’s two largest retailers—have AI-powered shopping assistants: Walmart’s Sparky and Amazon’s enhanced Alexa+. According to Walmart CEO John Furner in his April earnings call, shoppers who used Sparky spent roughly 35% more than those who did not, and the number of active weekly Sparky users more than doubled in the last quarter.

Walmart also reported new capabilities are being added to Sparky by the day, including automatic replenishment, meal planning and “more intelligent recommendations based on our inventory positioning, our prices, and our delivery speed capabilities,” said David Guggina, CEO of Walmart U.S.

Notably, the emphasis in his remarks is on Walmart’s inventory, prices and delivery speed—suggesting Sparky is optimized around the company’s priorities, not necessarily the customers’ stated needs.

This issue has caught the attention of the Groundwork Collaborative, a public policy advocacy group. In a The Capitol Forum op-ed, executive director Lindsay Owens wrote, “Troubling 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.”

“AI agents could be immensely useful,” she continued. “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.”

Retailers Can Take The Lead

Retailers have always prospered when they put the customers’ needs above their own. Walmart and Amazon are among the preeminent examples. But as Owens observed, even the leaders in AI development admit that they “do not understand how [their] own AI creations work.”

Before regulators and legislators take aim at their businesses, retailers have the opportunity to establish their own AI guardrails. That means ensuring that their proprietary AI agents, as well as the third-party agents they rely on, serve customers responsibly and with full transparency. Done right, retailers can lead this next era of AI-mediated shopping to benefit consumers and the long-term sustainable growth of their businesses.

See Also:

ForbesIs The Retail Industry Ready For The Booming “Chat” Economy?By Greg PetroForbesAI Ad Spending Will Reach $32 Billion In 2026, And Paid Search Teams Are Already Running ItBy Gabriel Alin Zainescu