Primaa

Deep learning image analysis is the entire offering. Cleo Breast and Cleo Skin are algorithms that detect and quantify biomarkers in whole slide images, with no viewer, scanner, or laboratory of the company's own. The deliberate choice to integrate into other vendors' platforms rather than build a workspace confirms the model: the company sells algorithms, not infrastructure.

The review requirement is stated, and the product page promises the software misses nothing. The Cleo Breast page says the tool is intended as an aid to diagnosis whose results must be reviewed by accredited doctors, and Cleo Skin carries the same sentence. The same Cleo Breast page says the software "detects every mitosis object and mitosis hotspots in the carcinoma regions" and "detects every potential calcification object", and a customer testimonial the company publishes there says the algorithm "ensures the quality and completeness of the evaluation by the pathologist". Those sentences make the product itself promise completeness: no mitosis or calcification goes undetected. Review by a doctor cannot answer that promise, because the reviewing doctor sees what the software flagged, not what it failed to flag, and mitosis counts drive tumor grade. Both products also prioritize cases into a smart worklist from an initial automated analysis, and the company publishes no threshold, override path or error handling for that ranking.

A stated review step does not offset a claim that the system misses nothing.

Product function is described concretely, including which biomarkers Cleo Breast addresses and named capabilities such as mitosis and lymph node metastasis detection, and the company states clinical development was conducted with European reference laboratories. What is not published is training data scale or composition, model architecture, or performance figures, and the completed results of the dermatopathology validation study were still pending at the time of review. Clear about what the products do, quiet about how well and on what basis.

A structurally short chain where Cleo runs in the laboratory, and an unnamed one where it runs in a partner's platform. Primaa builds its own models and states that Cleo does not collect personal data and that all data stay within the customer's own network, which, where it holds, means no outside party sits between the slide and the result. The same products are sold inside host platforms: PathAI's AISight lists them, and the company names PathPresenter, Proscia and Gestalt integrations. On those routes the host's infrastructure and terms carry the image, and Primaa publishes nothing about what the host passes to Cleo, where Cleo's processing runs in that setup, or which subprocessors apply.

A chain that is structurally short on the local route, with the host platform routes undescribed.

Published research by the company, short of independent validation. The company's evidence library lists a multicenter study of automatic invasive carcinoma detection on breast whole slide images (PLOS Digital Health), a 2025 evaluation of AI based mitosis detection with Cleo Breast, and a 2023 study of time savings with Cleo Breast, and the Cleo Breast page reports a clinical study with participating pathologists. In September 2026 Primaa and PathPresenter reported completing a joint validation of Cleo Skin inside PathPresenter's viewer, on a case mix of melanoma, nevus, squamous and basal cell carcinoma and benign lesions, describing faster and more consistent reporting without published figures. CE-IVDR marking implies a clinical evidence file reviewed by a notified body, which is not public. Customers quoted by name include pathologists at CHU Versailles and CHU Grenoble Alpes.

No study by a group unconnected to the company appears in the library, and the evidence measures detection and efficiency rather than what changes for the patient.

Published research short of independent validation.

A categorical commitment that patient data stays with the customer, carried in a product page footnote. The Cleo Breast and Cleo Skin pages state that Cleo does not collect any personal data, that the company's activities follow the GDPR, and that all data stay within the customer's own network. If that holds, the vendor never receives patient material, which answers retention and training use more firmly than any retention schedule could. A data protection officer is named in the website terms.

Three gaps remain. The statement is a footnote rather than a contractual instrument, and nothing explains how the boundary is enforced or whether it covers support access. It does not describe the routes where Cleo runs inside a partner's platform, such as PathAI's cloud hosted AISight, where images sit in the partner's environment. And the patient material used in the company's own development and clinical studies with French cancer institutes plainly reached the company by some route that is not described. Safety engineering, meaning how a wrong output is handled as a safety event, is not published.

What is missing is the documented detail behind a categorical promise.

A French company that does not invoke HIPAA, with a scope statement and no service document behind it. The product pages state that Cleo does not collect any personal data and that all data stay within the customer's own network, under the GDPR; the only legal document published is the website terms, which cover website data, and the privacy page carries cookie settings only. No HIPAA statement, business associate position or agreement appears, and no US entity or US customer is named, though the products are listed on US-based host platforms. Where a vendor never receives protected health information the business associate relationship may not arise, but the company does not make that argument, and the claim does not cover routes where Cleo runs in a partner's cloud.

The architecture is not set out in any service document, and Primaa does not say whether it will sign a business associate agreement covering support access and host platform routes.

Nothing independent is published about the security of the installed software. The site publishes terms and conditions covering the website and a privacy page that carries only cookie consent settings, and no ISO 27001 certificate, SOC 2 report, French HDS health data hosting certification, trust center, penetration testing statement or vulnerability reporting route is published. The company states that Cleo does not collect personal data and that all data stay within the customer's own network, which, if it holds, explains the absence of HDS, since a vendor that hosts no health data does not need it. It does not replace assurance over the software installed inside the laboratory, which is where the exposure sits under that architecture, or over routes where Cleo runs inside a partner's cloud platform. A data protection officer is named in the website terms, which is an accountability role, not a security credential.

Nothing independent published on security.

A European regulatory position stated, with research use only as the default elsewhere. Cleo Breast is described as a primary diagnostic solution with CE-IVDR marking, and a 10 September 2026 announcement says it obtained CE-IVDR marking for micrometastasis detection; the product page has not caught up and still marks micrometastasis detection as not CE marked. A March 2026 announcement says Cleo Skin received CE marking under the IVDR, which the company calls a world first for dermatopathology AI. The site's regulatory note states that unless otherwise stated all products are research use only and not for use in diagnostic procedures, and FDA's database lists no clearance for Primaa. On PathAI's AISight menu, which is PathAI's page rather than Primaa's, Cleo Skin appears as research use only, so its status depends on market and route.

A non US regime named, with research use only stated for the rest; no US clearance.

Generalization across sites is studied; performance across patients is not published. The company's evidence library lists a multicenter study of automatic invasive carcinoma detection on breast whole slide images (PLOS Digital Health) and a paper on standardized CycleGAN training for stain adaptation, both aimed at the staining and scanner variation between laboratories that most often breaks pathology models, and a clinical performance study with named French cancer institutes. None of that is a published breakdown of results by patient group, and dermatopathology AI carries a known concern about performance across skin tones that the company has not addressed publicly. No responsible AI statement or AI management system certification is published. Trade coverage at launch described a solution that refines its sensitivity as new data arrives; whether the deployed model is locked at a stated version is not published.

Work on site variation with no per group results published.

Results go to a doctor who must review them, and nothing stands behind the output. The Cleo Breast and Cleo Skin pages state that each is intended as an aid to diagnosis whose results must be reviewed by accredited doctors; lymph node metastases are shown as a heatmap, calcification tiles are displayed for the pathologist to check, and measurements of tumor size and margins are shown for verification. That is review before sign out, a mechanism for catching a wrong result the reviewing doctor can see.

The product pages give no error rate with a method and denominator, although the company's scientific evidence library lists a 2025 evaluation of its mitosis detection and a multicenter study of invasive carcinoma detection (PLOS Digital Health), whose figures the company does not carry onto those pages. No warranty, indemnity or remediation commitment attaches to a result, and the website terms disclaim responsibility for information on the site.

So the safeguard is the doctor's review, with nothing attached to the output itself.

Named host platforms, one listing verifiable, with direction and production use undocumented. Cleo is sold to run inside other companies' image management systems rather than its own workspace: PathAI's AISight algorithm menu lists Cleo Breast and Cleo Skin, Primaa announced integrations with PathPresenter (Cleo Skin, August 2026, with a joint validation completed in September 2026), and earlier material names Proscia and Gestalt environments and Sectra's viewer. The product pages say Cleo integrates fully into the laboratory workflow and with laboratory machines. No integration documentation, standard or production deployment by host platform is published, and whether results write back to a laboratory information system is not stated. For a pathology product the image management system and LIS, not the patient chart, are the correct surfaces.

Named systems with read access and results shown in the host viewer.

Residency is committed for the local route, with the host platform routes left open. The Cleo pages state that all data stay within the customer's own network and that Cleo integrates into the laboratory workflow and with laboratory machines, which places processing and slides on the laboratory's infrastructure. Cleo is also sold inside host image management platforms, including PathAI's cloud hosted AISight, PathPresenter, Proscia and Gestalt, and there the data rests wherever the host keeps it, which the local-network statement does not address. The commitment sits in a product page footnote rather than documented architecture, and no implementation requirements, supported systems list or timeline are published.

Options with residency stated for one route, and isolation and the processing path of the host routes left open.

No pricing is published. The multi platform distribution model means commercial terms likely vary by host environment, and the company discloses corporate milestones such as a pre Series A round rather than product economics. Whether a license covers Cleo Breast, Cleo Skin or both is not stated, a distinction that matters because the two differ in regulatory status.

Two specialties, one of them underserved. Breast pathology is the most crowded application in pathology AI, but dermatopathology has very little AI presence, and the company positions Cleo Skin as the first CE-IVDR certified tool in that specialty. A stated multi organ roadmap suggests intended expansion. Narrow today, but the dermatopathology position is differentiated rather than duplicative.