Independent FDA Regulatory Review of BloodGPT

Every company that builds AI for health data runs into the same question early. Does the software count as a medical device under FDA rules? The answer shapes everything downstream: how you market, how you sell, what you can claim, how long it takes to ship. We wanted a clear read on where BloodGPT stands, so we asked a firm that does this work for a living.

Hyman, Phelps & McNamara is a Washington firm focused on FDA and medical device regulation. We asked them to review both of our report products: the consumer report that a person receives after uploading a blood test, and the provider report that a clinician gives to a patient. The assessment they returned concluded there is a reasonable basis for the position that neither product is subject to FDA medical device regulation.

Two separate lines of reasoning support that conclusion, one for each product.

The consumer report fits the general wellness category. Under FDA's General Wellness: Policy for Low Risk Devices guidance, a product stays outside device regulation when it is intended to help people maintain a healthy lifestyle, presents low risk, and stays away from diagnosing or treating disease. BloodGPT reads a person's blood test, organizes the values, explains what the biomarkers mean in plain language, and tracks how they move over time. It does not tell anyone they have a condition. It points them toward a doctor. That framing is what keeps it inside the wellness category, and it is a framing we have to protect with every claim we make.

The provider report fits the non-device clinical decision support category. The 21st Century Cures Act carved out a class of software that supports clinicians without becoming a regulated device, on one central condition: the clinician must be able to independently review the basis for whatever the software shows. FDA sets out that condition in its Clinical Decision Support Software guidance. With BloodGPT, the physician keeps control over the report and its contents. The software highlights biomarkers outside normal ranges and adds educational context, but the clinician can see why each output appears and applies their own judgment to the final decision. That independent review is the line between a support tool and a device, and our architecture is built to sit on the right side of it.

The reason both conclusions hold comes down to how the product was designed, not to how it was described after the fact. BloodGPT normalizes each biomarker to LOINC, the international standard for reporting laboratory tests. Every value is checked against UCUM measurement standards. The platform benchmarks each reading against the laboratory's own reference ranges rather than a generic table. Every output traces back to its source, so the same input produces the same auditable result. A clinician can follow the path from raw lab value to displayed insight and verify each step. This is the glass box idea in practice, and it happens to be what regulatory review rewards.

A few honest caveats belong here. This is a legal assessment, not an FDA clearance or approval. FDA did not review or endorse anything. The conclusion reflects our current product and our current claims, and the firm was direct that changes to functionality or to marketing language could change the analysis. The memorandum also does not cover our laboratory-facing software that integrates with a lab's information management system, which sits in a different regulatory category, closer to FDA's definition of Software as a Medical Device. We treat the assessment as a baseline to maintain, not a box to check once.

For partners and labs evaluating BloodGPT, this changes a practical thing. The regulatory posture is no longer a matter of our own assertion. An independent FDA regulatory firm looked at the same product you would integrate and reached a documented conclusion. We will keep that analysis current as the product grows, and we will keep building so the answer stays the same.