“Made in USA” fraud isn’t just slipping past AI—it’s exploiting the AI governance gap at Walmart and Amazon

The core issue isn’t whether Walmart and Amazon can recognize potentially false “Made in USA” claims. It’s whether their AI shopping assistants translate that recognition into marketplace-level action (e.g., blocking, de-ranking, human escalation, or substantiation requests).

A Columbia Law School report titled “Made in America, Hidden by AI” argues that these systems demonstrate the capability to flag or respond to suspect origin claims while still allowing misleading listings to persist. Put differently: the same models and interfaces that understand origin fraud can be governed to ignore it—creating an enforcement vacuum that regulators are now explicitly trying to close.

Two strands matter for investors and policy-watchers.

1) AI governance behavior (capability vs. action): The Columbia report (primary source) describes Amazon’s and Walmart’s shopping chatbots/agents responding to “Made in USA” prompts in ways that demonstrate detection and evidence-gathering capability—while not implementing origin-claim enforcement mechanisms that would stop or systematically escalate misleading claims.

2) Regulatory posture shifting toward marketplaces: The FTC has already sent letters to platforms such as Amazon regarding misleading origin claims by third-party sellers, and later issued public warnings in 2026. Separately, an executive order signed March 13, 2026 targets “Made in America” truthfulness and calls out online marketplaces as an enforcement-relevant node.

The Columbia report’s mechanism is that model/assistant behavior changes when users explicitly specify “Made in USA” versus when they use query variations. This implies the system is not blind; it recognizes the issue class. The report further describes that platform filtering and refusal behaviors do not translate into a dedicated country-of-origin / “Made in USA” verification workflow for marketplace listings.

Two concrete technical inferences follow from the report’s described testing design:

• The system can be triggered reliably, so a compliance workflow could be turned on.
• But the workflow isn’t turned into enforcement, because listings and/or the assistant’s recommendations are not systematically constrained by substantiation standards.

That is the governance gap: detection without downstream compliance action.

The FTC’s enforcement logic is straightforward: under the FTC Act and the Made in USA framework, “Made in USA” claims must meet the “all or virtually all” standard. The FTC’s July 8, 2025 letter to Amazon explains what unqualified origin claims must mean and asks the platform to monitor, identify, and take corrective action regarding third-party sellers whose listings appear to make misleading domestic-origin claims.

Then, on July 6, 2026, the FTC issued additional public warnings that reinforce the agency’s continued focus on origin-fraud labeling.

Listed U.S. retailers with meaningful marketplace exposure can experience second-order effects as origin-fraud policing becomes a competitive variable. The direction depends on whether the company already has stricter controls for third-party claims.