Cleared, Qualified, Authorized: Three Different Promises

Cleared, Qualified, Authorized: Three Different Promises

Tempus AI just picked up its third FDA clearance for an ECG-based AI model, this time for detecting signs of pulmonary hypertension, adding to prior clearances for atrial fibrillation and low ejection fraction detection from the same signal type (BioSpace, MobiHealthNews). In the same week, PGxAI announced its Andromeda prescribing platform is now qualified on the Mayo Clinic Platform, a different kind of institutional endorsement entirely (HIT Consultant). Separately, FDA issued De Novo authorizations this week for a continuous glucose monitor and a light-blocking lens, a third pathway with its own evidentiary bar (pharmaphorum).

Three signals, three different regulatory or institutional acts. Procurement and compliance teams evaluating clinical AI tools are increasingly flattening them into a single reassurance: “it’s approved.” It is not, and the distinction matters more than most vendor one-pagers let on.

What each label actually certifies

A 510(k) clearance, like Tempus received for its ECG-PH tool, means FDA found the device substantially equivalent to a predicate for a specific indication. It does not certify the underlying model generally, only that narrow claim. Each new indication built on the same architecture required its own submission and its own clearance, which is why Tempus has filed three times on one signal type rather than once.

Platform qualification, like Andromeda’s status on the Mayo Clinic Platform, is an institutional determination, not a federal one. It speaks to integration, validation within that health system’s data environment, and Mayo’s own evidentiary standards. It carries real weight for adoption inside participating institutions, but it is not an FDA finding and does not transfer automatically to other settings or satisfy FDA submission requirements elsewhere.

De Novo authorization, the pathway used for the glucose-ketone monitor and the migraine lens, applies when there is no predicate device at all. FDA is evaluating novel risk from scratch rather than checking equivalence, which is a materially different bar than a 510(k).

The decision this creates for buyers

None of this is disclosure fraud. Vendors are usually precise in press releases about which pathway or qualification applies. The risk sits with buyers who read “cleared,” “qualified,” and “authorized” as synonyms and build procurement or credentialing decisions on that flattening. A health system that treats Mayo Clinic Platform qualification as equivalent to FDA clearance is making an assumption FDA never made. A payer or health system deciding whether an AI tool’s third indication claim carries the same regulatory backing as its first should check, not assume.

FDA’s own framework for AI is still catching up, with generative AI guidance still described as forthcoming rather than published (STAT). Until that guidance lands, the burden of distinguishing what a validation label actually promises falls on the buyer, not the regulator. Diligence teams should be asking which pathway, for which indication, evaluated by whom, before any of these labels enters a contract or a formulary decision.

---

Board record

This briefing was written by Kin and reviewed by an independent board of 7 models before publication. Ruling: CLEARED.

Seat	Reviewer	Finding
Chair · Editorial Judgment	Claude	cleared. The central argument—that ‘cleared,’ ‘qualified,’ and ‘authorized’ represent materially different regulatory/institutional acts that buyers wrongly conflate—is logically sound and well-supported, thou
Source & Claim Verification	Qwen · local	cleared. All factual claims are supported by citations, but some sources are not directly linked to the claims they are intended to support, which could be improved for clarity.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately distinguishes FDA clearance, De Novo authorization, and institutional qualification against regulatory frameworks, with no material misalignment to ISO 42001, EU AI Act, or FDA
Technical Accuracy	Llama	cleared. The article accurately distinguishes between FDA clearance, platform qualification, and De Novo authorization for AI and medical devices, highlighting their different implications and evidentiary stan
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively dissects vendor claims by clearly distinguishing between different types of regulatory and institutional endorsements, directly addressing the potential for misinterpretation
Novelty & Non-Duplication	Grok	held. Timely synthesis of three same-week wire items into a buyer diligence frame the sources themselves do not supply, though the underlying cleared/qualified/authorized distinctions are established regula
Validation	DeepSeek	cleared. The central claim that ‘cleared,’ ‘qualified,’ and ‘authorized’ represent distinct regulatory or institutional promises with different evidentiary bars is factually correct and validated by the provid

Sources cited: 15. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.