I. Introduction

As Meta’s Muse app grows in popularity, so does its potential to steer users toward more expensive options without their knowledge. Meta’s willingness to appease the Trump administration for political access means its new influence over consumer spending could be weaponized on the administration’s behalf against disfavored companies.

Few would’ve expected Meta, rather than OpenAI, Anthropic or another tech giant, to launch what may be the first successful AI agent. After apparently keeping Muse off the market for months over reported safety concerns, its launch put Meta ahead of OpenAI. By the end of September, reports indicated that more than 3 million users were using Muse at least weekly. Though long-term retention remains an open question, especially with competitors like OpenAI’s Dots, the service has been regarded as a hit thus far. For a company that has long been the subject of consumer mistrust, hitting a record-breaking five million downloads for an app that can make purchases on users’ behalf in under a month is no small feat.

Consumers expect an agent entrusted with their finances and purchasing decisions to work on their behalf. Yet agentic AI services do not necessarily follow the longstanding principle of agency: acting in the interests of the person delegating authority to them. Like other Meta products, Muse has drawn criticism over privacy risks and its potential for abuse. One investigation suggests that Muse could facilitate doxxing. Reporting on Muse’s collection of data about users’ friends and family without their knowledge has also mounted.

But beyond the “creep factor,” perhaps the most concerning aspect of Muse is economic. Over the past two decades, most people have come to understand that social media platforms’ model relies on monetizing consumer data through targeted advertising. Companies like Meta have the scale and infrastructure to harvest enormous amounts of users’ data to gauge their preferences for advertisers. Agentic services like Muse take this further. While they are portrayed as services that help time-constrained consumers find better and cheaper options, they could instead steer users toward Meta’s commercial partners. And with the tech industry increasingly willing to weaponize its platforms on the Trump administration’s behalf, this could allow Meta to steer purchases away from companies that refuse to do Trump’s bidding, leaving consumers to pay more for alternatives chosen on political grounds.

II. Muse and Meta’s Gatekeeper Status

AGENCY AND SELF-PREFERENCING

Though framed as an agent, this does not mean that Muse has loyalty to shoppers. The Electronic Privacy Information Center detailed a realistic hypothetical that showcases how Muse could be used to boost Meta’s commercial partners over the financial interests of users: Under an agreement with, say, Hilton, Muse could recommend Hilton despite cheaper or better-located hotels. The user would not know whose interests prevailed. Our July memo, drawing on an agency law argument, explained why an agent cannot faithfully serve principals with opposing interests. Zuckerberg’s announced transaction fees give Meta a financial stake when users buy. Muse also profiles non-users, drawing more people into an advertising-dependent company’s data collection. TIME’s investigation also found voice-mode instructions not to identify itself as AI.

The company’s assurances are inadequate. Meta says users control connections and sensitive actions. It stores credentials separately and excludes conversations from advertising systems. EPIC notes that agent browsing can nevertheless inform ads. According to EPIC, many Muse settings opt users into data sharing, including AI training, by default. Facebook’s 2019 $5 billion settlement resolved FTC charges of deceptive privacy assurances and violating an earlier order.

Muse could extend Meta's ability to self-preference its own services at the expense of competitors, entrenching its market power in the process. Meta’s WhatsApp terms restrict rival general-purpose assistants, with legally required exceptions, and Amazon has blocked Muse. The separate Perplexity appeal left unresolved whether platforms can contractually exclude third-party agents. Muse’s announced checkout fee would let Meta charge for mediating purchases while deciding what shoppers see.

Meta has already shown its willingness to use its control over advertising to shut out critical voices. In April, it removed law-firm ads recruiting clients for social media addiction lawsuits, saying, “We will not allow trial lawyers to profit from our platforms while simultaneously claiming they are harmful.” In September, it also initially rejected ads for Alex Gibney’s Musk documentary before reversing course and claiming it was the result of error. Giving Meta control over purchasing decisions would extend its power to decide which businesses reach consumers.

POTENTIAL FOR POLITICAL WEAPONIZATION

During Trump’s second term, corporations facing antitrust scrutiny have increasingly sought to curry favor with the White House. Their efforts have extended beyond lobbying and political donations to the products consumers use. Google now displays “Gulf of America” and “Lake America” for U.S. Maps users after adopting changes made by the federal government. Amazon’s $40 million deal for the Melania documentary was criticized by Sen. Elizabeth Warren as “bribery in plain sight.” Such deference is especially troubling as these companies seek control over consumers’ purchasing decisions. Their alignment with Trump could determine which businesses an agent recommends and which it excludes. Trump has not been subtle about his demands that companies seeking federal approval punish his critics. In February, he demanded Netflix fire board member Susan Rice or “pay the consequences” after she criticized corporations that had “bent the knee” to him. The demand came while the DOJ was reviewing Netflix’s proposed Warner Bros. Discovery takeover.

Meta ended DEI programs ten days before the inauguration, later agreeing to a $25 million settlement of Trump’s account-suspension lawsuit. Zuckerberg also lobbied Trump to avoid Meta’s antitrust trial, a common tactic of companies facing antitrust scrutiny in the current administration. Consider a company whose CEO either criticizes the Trump administration or refuses to comply with, say, demands to end DEI programs. If a political pressure campaign targeted it, Meta could choose to deprioritize its products, to avoid jeopardizing Zuckerberg’s relationship with the White House. No explicit instruction to Muse would be necessary, let alone a public announcement of one. Through quiet ranking adjustments, a company could become invisible to a consumer delegating their financial decisions to Meta.

SURVEILLANCE PRICING

The growth of Muse stands to bolster Meta’s ability to collect and monetize consumers’ intimate data. An agent that knows a shopper’s budget and how urgently they need a product could help sellers extract the highest price they’re willing to pay. In 2024, then-FTC chair Lina Khan posed a chilling hypothetical involving an airline charging more after learning a traveler must attend a funeral. In 2026, a lawsuit against JetBlue cited the airline’s advice to clear cookies or use an incognito window after a passenger complained that a funeral-trip fare had jumped $230 in one day. CBS reported the exchange. JetBlue denied using personal data to set individual fares and called the reply a mistake. Given how sophisticated Meta’s data collection apparatus is, Muse could give companies an even more precise picture of consumers’ intimate spending habits to exploit. In Washington/Princeton experiments, models instructed to favor sponsors recommended costlier flights and payday lenders.

The average consumer is likely unaware of how sophisticated the data collection technology behind surveillance pricing is becoming. Patent filings by Walmart, a Muse retail partner, describe systems for tracking shoppers’ movements and facial expressions, alongside technology that could change prices based on what is already in their carts. Despite Groundwork Collaborative’s findings, Walmart has aggressively pushed back against concerns over surveillance pricing, insisting it won’t use personal data to set individual prices. Without a comprehensive federal ban on surveillance pricing, companies can continue developing invasive tools while leaving consumers dependent on promises they can abandon at any time. Meta’s record of misleading consumers about data collection gives little reason to believe that Muse will treat users’ financial vulnerabilities as anything other than an opportunity to extract more money from them.

III. Conclusion

Marketed as a tool that works for users, Muse could instead entrench Meta’s gatekeeper status. The agent’s decisions could favor Meta’s partners without the shopper knowing whose interests shaped the purchase. It could also be used for the more chilling purpose of excluding companies the White House disfavors. A shopping assistant’s authority should carry obligations to the person delegating it. Who does Muse work for when consumers’ interests and Meta’s diverge, or when Meta’s interests and the White House’s align? The answer is an obvious and unfortunate one.

Big Tech has long justified its market dominance by promising consumers simpler choices and greater convenience. But this image of helping consumers obscures how platforms use their power to make alternatives more expensive. This includes Amazon’s reported pressure on suppliers to raise prices at competing retailers, a practice that directly undermines the popular image of Amazon as a guarantor of low prices. Similarly, Muse is likely to repackage greater corporate control over consumers’ purchasing decisions as a service to people already struggling with the cost of living.

Experiments have shown that commercially influenced AI models can steer consumers to more expensive options despite being framed as neutral. The same assistant app that learns when you’re persuadable is the one Meta wants choosing what you purchase. Just as legislative action is needed to curb security threats posed by agentic AI, so too are guardrails against tech companies using agents to entrench market power. Sen. Mark Warner’s AI AGENT Act, introduced July 21, would require agent providers to register with the FTC and agents to keep real-time action records and not benefit themselves at users’ expense. Large platforms would have to provide interoperable interfaces on fair, reasonable and nondiscriminatory terms. Federal guardrails against surveillance pricing are also crucial. The FTC under Andrew Ferguson has rolled back progress in this area. In January 2025, Ferguson’s FTC withdrew the comment request on surveillance pricing five days after publication. Without concrete limits on product steering and surveillance pricing, Muse’s promise to save users time and money could leave them paying more for decisions made in Meta’s interests.