US Baby-Food Heavy-Metal Laws: States Move Ahead of FDA

Abstract

Virginia, Illinois and Vermont now provide official state-law records for baby-food heavy-metal testing, disclosure or sale requirements, while FDA continues to operate through federal guidance, monitoring results and planned action-level work. These layers are related but not interchangeable. A state can require testing records, consumer-facing disclosure or a QR pathway even when the underlying federal action level is nonbinding guidance. FDA’s 2026 infant-formula testing release and program priorities provide important evidence and policy direction, but they do not themselves create a new binding federal maximum level. This analysis compares the legal mechanisms, explains why product definitions and effective dates must be resolved state by state, and maps the consequences for a certification evidence package without converting state duties into universal HMTc thresholds.

Federal Guidance Is Not the Whole Compliance Map

FDA’s contaminant work has national reach, but much of the current baby-food framework operates through guidance and action levels rather than one federal statute that preempts every state operational duty. FDA’s final lead guidance can inform agency enforcement, yet guidance does not automatically erase a state’s authority to require testing, disclosure or conditions of sale. FDA’s 2026 program priorities identify work toward cadmium and inorganic-arsenic action levels; that is a work plan, not a completed binding rule ¹. The compliance question therefore changes by layer. The federal question may be whether a product aligns with an FDA action level and the adulteration framework. The state question may be whether the manufacturer tested the right lot at the required cadence, retained the right record, used the required disclosure channel, covered the statutory product definition, or met a state sale condition. Passing one question does not answer the other.

Virginia, Illinois and Vermont Use Different Legal Levers

Virginia’s official code creates a Baby Food Protection Act section with testing and consumer-access provisions tied to covered baby food ². Illinois’s official legislative record shows SB0073 becoming a Public Act with a 1 January 2026 effective date and a structure that combines sale requirements with testing obligations ³. Vermont’s H.536 record became Act 110 and requires close attention to product definitions, including when infant formula is within the pathway ⁴. The useful comparison is not a red-or-green state map. It is a matrix with at least six fields: covered product definition, covered analytes, concentration or sale condition, testing cadence and laboratory requirements, disclosure mechanism, and effective or transition date. A company that normalizes only metal numbers will miss the operational duties most likely to produce a state compliance failure.

FDA Testing Adds Evidence, Not a New Limit

FDA’s 2026 infant-formula program is significant because the agency released its largest testing effort for infant formula, with an official product-results surface and a separate explanation of the program ⁵. Those results can improve occurrence understanding, identify analytes that need closer surveillance and inform future federal policy. They do not, by publication alone, create a new maximum concentration that every lot must meet. For evidence-to-action purposes, monitoring and law occupy different columns. The dataset belongs in occurrence and risk review. A final action level belongs in the federal regulatory comparison with its correct legal weight. A state law belongs in the operational matrix. Keeping those columns separate prevents a monitoring result from being described as a ban and prevents a state QR rule from being misdescribed as a federal concentration ceiling.

The Certification Evidence Package Must Be State-Aware

A defensible certification package can support state compliance, but certification is not a substitute for legal review. At minimum, the package should preserve lot identity, sample date, product form and basis, analyte species, method and detection capability, laboratory identity, result, release decision and the public disclosure artifact when one is required. It should also preserve the product-definition determination that explains why the SKU falls inside or outside each state act. The HMTc standard remains a separate object. If a state creates a lower binding limit for an in-scope product and that limit is loaded into the normal regulation compiler, the published standard can be reviewed through governance. Testing frequency, QR publication and record-retention duties can be added to an evidence checklist without pretending they are concentration thresholds. Nothing in this analysis silently changes a standard; it identifies which evidence a certified brand will need in each market.

A Practical Multi-State Control

Brands should maintain one controlled state-law register per SKU family, not one national baby-food checklist. The register should cite the official statute, show current status and date, record the operative product definition, list metals and method requirements, identify the disclosure URL or QR destination, and name the owner who rechecks the rule. It should point to the certification record rather than copying test results into unmanaged spreadsheets. The register also needs a negative statement: what the rule does not do. Virginia does not become federal law; Vermont’s product definition does not automatically apply in Illinois; and FDA’s planned cadmium or inorganic-arsenic work is not binding until the agency issues an instrument with legal effect. That negative control prevents a fast-moving state landscape from becoming an internally inconsistent set of claims.

Frequently asked questions

References

Works cited in this analysis’s text, in first-appearance order. This is not the full evidence base for the finding; it is what the prose above draws on. The complete occurrence record and per-source pages live on the Heavy Metal Index: this analysis links to them rather than re-hosting them. works without a verified DOI, and primary legal instruments and published standards, are named but not linked.

Cite this analysis