AI is the first line of defense against numerous forms of marketplace deceit in the speedily evolving realm of e-commerce. However, a recent study from Columbia Law School’s Center for Law and the Economy shows that there is a troubling disconnect between technology and action at the corporate level. The study is the first from the center that Lina Khan, the former chair of the Federal Trade Commission (FTC), has established, and it shows that the “Made in USA” claims are not always accurate, but that Amazon’s shopping assistant Alexa, and Walmart’s shopping assistant Sparky, are able to tell the difference reliably—and are on purpose not doing so to remove or mark misleading listings.
This discovery is much more than just consumer protection. The Made in USA logo is a big deal when customers are searching for products to purchase online, as it not only signals patriotism, but also can indicate that they are willing to pay premium prices for products made in the United States. The Federal Trade Commission regulations governing the use of the term “Made in USA” say that the product must be “all or virtually all” U.S.-made, meaning nearly all its important components and manufacturing processes must be American. The FTC’s expectations are obvious, and last year the agency specifically asked Amazon and Walmart to be more vigilant in policing third-party sellers’ “Made in USA” labeling to ensure there is accurate information.
The key is that the AI programs that these companies have already implemented could easily detect fraudulent claims. The study’s authors discovered that in cases where there is a clear “Made in USA” claim, Alexa and Sparky are able to spot inconsistencies by cross-referencing conflicting information, which might be found in product descriptions, manufacturing location data or component sourcing information, all within the product listing and in the same language. This technical capability is very powerful in the realm of using AI to protect consumers, but it is not widely used for enforcement.
After pressuring the AI assistants on why the companies are not doing enough to counter deceptive labelling, the answers came back with a business strategy take on, not a technical constraint. The study reports that Walmart’s Sparky told researchers that the FTC usually regulates the “Made in USA” claims of manufacturers, not retailers. Perhaps most strikingly, the chatbot shared the following: “That’s a business calculation, not a legal justification! The finding of this study – this response – captures the essence of the matter: these companies have made a conscious choice to prefer other considerations over enforcement.
The same was true on the Amazon side of the equation. Alexa “admitted this is a serious problem and is ‘easier to do nothing’ until the harm to U.S.-made brands has become ‘financial, regulatory, or reputational cost’ to Amazon in particular,” she said when confronted with the issue of continuing to label products as “Made in USA” when they are not. The stark admission by the AI itself brings up a key balance between what the technology can do and what is being gained by using it.
WalMart didn’t immediately comment on the conclusions of the study. In the meantime, an Amazon spokesperson said that “country-of-origin information, when available, is currently provided on product detail pages and we are making constant efforts to make this information even more readily available for customers to use in our Alexa for shopping features. This recognition of the need for transparency in country of origin is noteworthy, but it preempts the question about enforcement that the research raised.
The study comes from a very interesting institutional setting. Since Khan’s departure from the FTC, the Columbia Law School’s Center for Law and the Economy has emerged, a part of which has been led by Khan.Since Khan left the FTC, the Columbia Law School’s Center for Law and the Economy has been established, and Khan has been one of its leaders, with notable victories such as a lawsuit against Amazon that alleged the company was exploiting its online retail monopoly. That case is pending and will further complicate Amazon’s public posture on marketplace fairness and adherence to regulations.
This has considerable impact on the consumer. Those who specifically look for American-made goods for domestic manufacturing, quality assurance or for feeling good about the purchase may be buying goods that fail to meet the standards of the third group. Why isn’t this technology deployed, given that it is readily available, is a major question that arises, and one that raises fundamental questions about the link between corporate responsibility and consumer protection in the digital age.
The business incentives that Sparky identified shed light on the seeming paradox. As “Made in USA” is claimed on most products, any such enforcement would necessitate significant resources for major retailers running large third-party marketplaces, would be likely to impact their relationship with sellers and could decrease selections or lead to premium pricing of the actual “Made in USA” product. In addition, the issue of liability arises from the fact that platforms routinely police and remove listings, which could be interpreted as assuming more responsibility for each listing made by third parties and thus creating more potential for liability.
But the study’s results resute that it’s not impossible or especially hard to enforce. The AI systems already prove their ability to identify fraudulent claims, indicating that it is not a technological hurdle, but rather corporate will. This leaves a less-than-welcome situation where the mechanisms to safeguard consumers are in place but being thrown to the side.
