Glass Health

The reasoning engine is the product. Ranked differentials, cited clinical question answering across a 38 million article index, and drafted assessment and plan are all model outputs, and the Developer API sells that intelligence directly as a component to other platforms.

The strongest structural answer to the hardest problem in clinical AI. Every response carries inline citations to source literature and guidelines, so a clinician can verify any individual recommendation against the evidence rather than trusting the output. The company insists on the product being decision support requiring physician judgment on every output, and it declines to position it otherwise. Verifiable reasoning plus a stated ceiling on autonomy is the right posture for differential generation.

Unusually specific about what the system reasons over: an index of more than 38 million peer reviewed articles, consensus guidelines, and FDA information covering over 154,000 drug compounds, with inline citations exposing the sources of any given answer. For a decision support product, disclosing the evidence corpus and making every output traceable to it is the substantive form of transparency, more useful than naming a base model.

The evidence corpus is disclosed with unusual specificity and the model layer is not. The system is described as reasoning over an index of more than 38 million peer reviewed articles, consensus guidelines and regulatory drug information covering more than 154,000 compounds, with inline citations exposing the sources behind any given answer.

For a decision support product that is a real answer to half of this axis, because the data contributing to the output is both named by category and traceable per answer, which is more than most vendors disclose about any input. What is absent is the other half. No foundation model provider, model class or version is named, no hosting arrangement is published, and no sub processor list was located.

Nor is it stated how the answer differs between the free individual tier and an enterprise agreement, and that distinction matters here more than at most vendors because the product also performs ambient scribing and draws patient and visit context into subsequent questions, so a clinician on a free tier may be sending encounter content into a chain governed by consumer terms. Ask which model provider sits behind the reasoning layer, for a sub processor list, and for the chain and terms applicable to the tier you are on.

Evaluated in the Harvard and Stanford ARISE NOHARM benchmark, a physician validated clinical safety evaluation, where the platform ranked among the top medical AI systems. Independent safety benchmarking by academic institutions is a materially stronger evidence position than vendor accuracy claims, and it is the appropriate evidence standard for a product that proposes differentials. Independent testing also documents the limits, noting reasoning is weaker on complex multi system cases than on common presentations.

This product handles protected health information at depth and at volume, which is why the silence matters. Ambient scribing means recorded patient encounters, the platform draws patient and visit context into subsequent questions, and the API accepts production data of this kind under a business associate agreement. So the stewardship question is live in a way it is not for a pure reference product. What is published is the compliance frame and essentially nothing beyond it.

The HIPAA and agreement position is stated clearly, and answers carry in text citations and structured reference sections so a clinician can audit the evidence behind an output. Nothing was located on the questions this axis actually turns on: whether encounter audio or transcripts are retained and for how long, whether customer content or protected health information is used to train or improve models, what the deletion path is, or how the answer differs between the free individual tier and an enterprise agreement.

That last distinction matters more here than at most vendors, because the same product is sold both ways, and a clinician on a free tier recording a patient encounter is exactly the bottom up adoption pattern that has caused institutional compliance problems elsewhere in this category. The terms of service and privacy policy were not opened in this review, so this assessment could change.

The mechanism is the notable part here, and it is the most operationally concrete route to a business associate agreement in this segment. The developer documentation states that Glass is end to end HIPAA compliant and that teams can review and accept such an agreement in API settings before sending production protected health information.

An agreement executed as a self serve step inside the product, gating transmission of that data, is a different and better thing than one promised in a sales conversation. Set against the rest of the segment the contrast is sharp: one competitor publishes a standard agreement but a named health system states publicly that it could not reach acceptable terms, another mentions HIPAA nowhere in its privacy notice or terms, a third's dedicated compliance page names no framework at all, and a fourth's posture could not be established from public material.

Held below a higher grade for three specific reasons. The statement sits in the developer API documentation, so it is not established that the same self serve path governs the clinician facing product. End to end HIPAA compliant is a strong self assertion with no attestation located behind it. And no terms, tier, subprocessor list or execution detail was retrieved. Whether the in product agreement covers the main application or only API customers is the highest value question on this record.

A third search, run through the developer and legal surfaces rather than the marketing pages, again returned no SOC 2 report of either type, no HITRUST certification, no ISO 27001, no penetration testing statement and no public trust centre for this vendor. The absence is now tested rather than assumed, and two things make it matter more here than it would elsewhere.

The segment comparison is unflattering. Clinical reference and ambient documentation are categories where an independently examined security posture has become the procurement baseline. Competitors publish one, several surfacing it in a persistent site footer or through a full trust platform with an annual external penetration test and a vulnerability disclosure programme. Measured against its own segment, an absence is conspicuous rather than merely unstated.

The second point is specific to this record. The company sells a developer interface delivering clinical question answering, record summarisation, differential generation and ambient scribing to other companies building clinical products. A developer buyer does not simply trust this vendor, it inherits this vendor's posture into its own product and passes it downstream to health systems. Subprocessors, retention, logging and hosting regions are more load bearing for an interface business than for a direct application, not less.

The sharpest point is the company's own. Glass Health publishes a buyer's guide to evaluating healthcare AI interfaces which tells readers to establish business associate agreement scope, eligible services, retention, logging, regions and subprocessors before committing. That is the right list. The company does not publicly answer it for itself.

The grade reflects what a counterparty can verify before entering a sales conversation, not a judgement that controls are absent. A company holding production protected health information under business associate agreements will have security documentation and enterprise buyers will have seen it under non disclosure. Ask for the list this vendor itself published.

No FDA clearance, none claimed and none apparently sought. The clinical decision support exclusion position is supported by the product's own architecture rather than merely asserted: outputs carry in text citations and structured reference sections, the company states the product is decision support requiring physician judgment on every output, and independent testing is acknowledged rather than suppressed.

But this record sits closer to the regulatory line than others in this segment, and that should be said plainly. Comparable products answer questions or surface protocols. This one generates a ranked differential diagnosis with suggested workups from a case description.

Producing an ordered list of candidate diagnoses is a materially different act from retrieving what the literature says, and the exclusion turns on whether the clinician can independently review the basis for the recommendation, which is harder when the output is a ranking whose ordering logic is not itself shown. That is an observation rather than a legal conclusion.

Two things are worth asking the company directly: its regulatory rationale for differential generation specifically rather than for the platform in general, and how that position is maintained given that its developer API lets third parties embed the same differential engine in products the company does not control.

This grade rests on something rare in this index: the vendor submitted itself to an external, third party, physician validated safety benchmark and the result is public. The Harvard and Stanford ARISE NOHARM evaluation placed the platform among the top medical AI systems, as recorded at the time of assessment. Submitting to an evaluation you do not control is categorically stronger than publishing principles about evaluation.

The company also states its outputs are evaluated by clinicians and benchmarked on clinical use cases with healthcare specific safety review, and it acknowledges publicly that reasoning is strongest on common presentations and weaker on complex multi system cases. A vendor that publishes where its product is weaker is doing something most competitors do not.

What is absent keeps it below the top grade: no model card, no bias or demographic subgroup analysis of any kind, no error taxonomy, no performance breakdown by specialty or presentation type, and no published error rate. The contrast with the leading subscription reference product in this category is the sharpest governance pairing in the segment, and both sit at the same level for opposite reasons.

This vendor submitted to someone else's measurement and published the result but offers no framework of its own; the other published a framework for evaluating clinical AI and withheld every result from it. One has the measurement without the framework, the other the framework without the measurement.

The mechanism here is the right one for a decision support product and it is well executed. Answers carry inline citations and structured reference sections, so a clinician can follow any statement back to the article, guideline or regulatory record it came from and judge the source directly.

For this product category that is more useful than an accuracy percentage would be, because a clinician who can see the underlying evidence is not relying on the system's judgement at all, they are using it as a retrieval layer over material they can assess themselves. It is also the mechanism that makes an error contestable at the moment of use rather than discoverable afterwards. Held at C because nothing measures how often the system is right and nothing stands behind it.

No accuracy figure, error rate or evaluation methodology was located, no warranty, indemnity or remediation commitment exists, and citation traceability does not address the failure mode that matters most in retrieval based systems, which is a correct citation attached to a claim the source does not actually support.

One commercial detail deserves attention: the same product is sold to individual clinicians and to institutions, and the terms of service and privacy policy were not opened in this review, so what an individual user agrees to is unestablished. Ask whether the citation is validated against the claim it supports, and read the terms for the tier you are on.

Four electronic health records are named, Epic, eClinicalWorks, athenahealth and Elation, but the language is assisted workflows rather than integration, the capability is gated to the top paid tier, and the company's own comparison material tells teams to confirm the current implementation path directly with Glass. A vendor hedging its own integration claim is the hedge this axis exists to catch.

No Epic Showroom or marketplace listing was located, no FHIR or SMART on FHIR conformance statement, no named access route inside the chart, and no published integration effort. The direct comparison within this segment is unforgiving, since a competitor holds a live Epic Showroom listing and an athenahealth Marketplace listing, writes notes and pends orders back into the chart, and publishes an Epic build at under 20 hours.

The real interoperability strength here is a different shape and is credited rather than ignored: the Glass Developer API exposes the whole stack, covering clinical question answering, chart summarisation, differential generation, care planning and scribing, so other platforms can embed the clinical intelligence directly.

That is a genuine and unusual interoperability surface, but it is an integration path for other software vendors rather than depth inside the electronic health record a clinician actually works in, which is what this axis measures.

The delivery model is clear and unusually self serve for clinical software: a web platform with a free tier, a published plan ladder with direct signup, and team and enterprise tiers above it, plus the developer API as a separate deployment surface for embedding. A clinician can be using this product within minutes without an institutional purchase, which is a real and deliberate distribution choice. What is missing is everything this axis asks about on the institutional side.

No data residency statement of any kind was located, no region named and no hosting provider disclosed, which matters because the product accepts production protected health information and records patient encounters. Competitors in this segment name their cloud providers or state their country of storage outright. No named health system deployment was retrieved, no implementation shape, no uptime commitment and no support model.

The only adoption figure available dates from 2023 and is now three years old, alongside an enterprise offering described as a pilot at that time. A self serve product with no published institutional deployment evidence is a different procurement risk from a gated one, and establishing where recorded encounters physically live is worth doing before piloting.

Freemium with a genuinely usable free tier for individuals covering limited scribing and decision support, and a Pro plan reported around $90 per month, with enterprise and Developer API pricing quoted through sales. A clinician can evaluate and adopt without procurement. Held back from A because the Pro figure comes from third party review rather than a published rate card, and API pricing is entirely undisclosed.

Coverage is broad across both the documentation and the reasoning sides, and it is enumerated rather than implied. Note types span clinic notes, history and physical notes, progress notes, discharge summaries and patient handouts. Reasoning outputs span ranked differential diagnosis with suggested workups, problem oriented assessment and plan drafting, and clinical question answering.

The grounding corpus is stated as more than 38 million peer reviewed articles plus consensus guidelines and FDA information covering over 154,000 drug compounds, which is breadth comparable to the largest literature indexes in this segment and far wider than any curated corpus product. Buyer segments span independent practice, medical group and academic medical centre, so the product is not scoped to one setting size.

Held below a higher grade on a limit the company itself acknowledges, which is why it is credited rather than penalised twice: independent testing notes that reasoning is strongest on common presentations and weaker on complex multi system cases. That is the same shape as findings elsewhere in this segment and means breadth of coverage is established while depth at the hard end is not.

No specialty specific validation, performance breakdown by presentation type, or named specialty society content partnership was located, the last being the characteristic third party credential in this segment and one competitors hold in numbers.