Indica Labs

The cleared product is a platform, not an algorithm. HALO AP Dx is an enterprise digital pathology workspace whose FDA clearance covers primary diagnosis, meaning viewing and diagnosing from digital slides, and that clearance rests on the platform plus scanner pairing rather than on AI performance.

AI is a real and central part of the offering, including hosted third party algorithms and the company's own image analysis, but a laboratory could deploy the cleared platform for digital primary diagnosis and use no AI at all. That makes HALO an established platform that adds AI, which is not the same as a product that is AI.

The platform's cleared function is to let a pathologist make the primary diagnosis on a digital slide, so the human is the diagnostic actor by design. Algorithms accessed through the associated store are assistive and, as the company notes, carry their own separate regulatory classifications, which is an important disclosure: platform clearance does not extend to the algorithms running on it.

The regulatory scope is stated with real precision, including clearance numbers tied to named scanner hardware, and the company is consistently explicit about which product is cleared and which is research use only. What is not published is model level detail on its own AI, training data, or validation, because the disclosure emphasis sits on platform clearance and integration rather than algorithm performance. Transparent about scope, quiet about models.

The host is named and the tenancy is the customer's, with the hosted algorithm layer undescribed. Indica's Cloud Services run HALO on Amazon Web Services inside an AWS account the customer controls, so the cloud provider is named and the data sits in the customer's own tenancy; laboratories can also run HALO on their own infrastructure. Indica builds its own AI (Breast IHC AI, Lung PD-L1 AI, Prostate AI and the HALO AI trainable classifier), so no outside model provider sits behind those. The platform also hosts third party AI, with a strategic alliance with Lunit named, and third party algorithms run inside HALO AP.

No subprocessor list is published, and nothing says what data a hosted third party algorithm receives, where it processes it, or whether that vendor may retain or train on it.

The host named and the model layer in house, with the hosted partners' terms missing.

Cleared and widely deployed, with little published about how well the AI performs. The 510(k) clearances for HALO AP Dx rest on validation of the viewing platform for primary diagnosis, and the company names deployments at the National Cancer Institute, the University of Washington (40 percent AWS cost savings in four months, a vendor figure), Nottingham University Hospitals, the University of Bern and Source BioScience. The AI product pages describe their benefit in words (standardized scoring, reduced interobserver variability without compromising accuracy) and publish no study, method or figure for Breast IHC AI, Lung PD-L1 AI or Prostate AI.

Named customers and vendor figures without method for the AI products.

General compliance assurances, with the customer's own cloud account doing the real work, and nothing on retention or training. The product pages state that HALO AP provides built-in compliance with FDA 21 CFR Part 11 (electronic records and audit trails), HIPAA and GDPR. Under Cloud Services the deployment sits in an AWS account the customer controls, which keeps slide images in the customer's tenancy rather than a vendor environment. The privacy policy (effective 15 May 2026) covers the website, webinars and marketing, not the platform or slide images.

No retention period, deletion position, de-identification practice or statement on whether customer images train Indica's models or third party hosted models is published, and nothing describes what a hosted third party algorithm receives. Safety engineering, meaning how a wrong model output is handled as a safety event, is not published.

General assurances that do not answer retention, training use or de identification.

HIPAA compliance is claimed as a built-in property, with no business associate position or service document behind it. The product pages state that HALO AP provides built-in compliance with HIPAA alongside FDA 21 CFR Part 11 and GDPR, and the platform is sold to US clinical laboratories through HALO AP Dx and under a managed cloud service in which Indica's team runs the customer's AWS environment. No business associate status is stated, no agreement is named or published, and the only privacy policy (effective 15 May 2026) covers the website, webinars and marketing rather than the platform or patient data.

Compliance claimed without a business associate position or a service privacy document.

An outside check of the cloud deployment, and no security attestation. The Cloud Services page says Indica's AWS deployments of HALO are validated by AWS, with Amazon CloudWatch monitoring and an AWS certified team running them, and the company is an AWS Partner Network member. The product pages state built in compliance with 21 CFR Part 11 electronic record controls, HIPAA and GDPR, and the company holds ISO 13485:2016, a medical device quality standard rather than a security one. No ISO 27001 certificate, SOC 2 report, HITRUST certification, trust center or penetration testing statement is published.

The controls have an outside check of the cloud architecture behind them, but none of it is attested against a recognized security framework.

The regulatory position is identifiable, and the company keeps the AI out of the cleared label. FDA's database lists two 510(k)s for HALO AP Dx: K232833 (7 May 2024, product code PSY), with the Hamamatsu NanoZoomer S360MD, and K252762 (25 November 2025, QKQ), adding the Leica Biosystems Aperio GT 450 DX. The HALO AP Dx page states that it is FDA cleared for primary diagnosis in the United States and approved with the Aperio GT 450 DX and named displays for SVS and DICOM files; the clearance is for the platform, not for an algorithm.

The AI products carry their own stated positions. HALO AP, the platform that runs Indica's AI and third party AI, is CE-IVDR marked for in vitro diagnostic use in Europe, the UK and Switzerland and is stated to be research use only in the United States and not FDA cleared for clinical diagnostic use. Breast IHC AI 2.0 is stated to be research use only and not intended for clinical diagnostic use, and Prostate AI is described as CE-IVD marked. No AI product is called FDA cleared.

The company answers the platform's governance question with product rather than policy. Where customers build their own models, the question is who establishes that a customer built model works. HALO AI ships validation workflows as a feature: users train classifiers, run them against a validation set, and see regions of interest marked for true positives, false positives and false negatives, with the process centralized rather than assembled by hand. That is material for the laboratory's own review of the model, placed with the party that built it.

Three gaps remain. The tooling shows whether a customer's classifier works on that customer's validation set, not how Indica's own AI products perform across patient groups, scanners or stains, and no subgroup results are published for them. There is no responsible AI statement or AI management system certification. And for hosted third party algorithms, nothing published says who validates one in a given laboratory, who monitors it after go live, or where accountability sits when it is wrong.

What exists supports the customer's own evaluation of its models and stops short of subgroup results.

Mechanisms let a pathologist check and trace results, and nothing stands behind the output. Breast IHC AI reports the percentage of positive cells and the biomarker score at slide level together with image analysis masks the pathologist can inspect in HALO AP, and the platform claims built-in 21 CFR Part 11 compliance, meaning audit trails and record integrity controls that let a laboratory reconstruct who changed what. The pathologist signs out every case, and the AI products are research use only in the United States.

No error rate with a method and denominator is published for Indica's AI products on the product pages; the clearances concern the viewing platform, whose performance is not an error rate of an AI function. No warranty, indemnity or remediation commitment attaches to an AI result.

Review and audit mechanisms with nothing attached to the output.

Two way LIS integration is described, with the systems themselves unnamed. The HALO AP Dx page says the platform assembles cases from barcodes and LIS or LIMS data, keeps worklists updated from the LIS, generates the pathology report and sends it automatically to the LIS through an interface, and offers API-first connectivity; it is DICOM compliant and cleared for SVS and DICOM files from the Leica Aperio GT 450 DX. It directs buyers to email for the list of existing LIS and LIMS integrations rather than publishing it. Voicebrook's VoiceOver PRO integrates for voice navigation and reporting, and the platform hosts third party AI, including through a strategic alliance with Lunit. Deployments are named at the National Cancer Institute, the University of Washington, Nottingham University Hospitals and Source BioScience.

Standards support and a described bidirectional LIS interface with named deployments, short of named LIS integrations a buyer can verify.

Options stated and the cloud model is unusually specific about ownership, with residency and the management path left open. HALO software can be installed by the laboratory or run under Indica's Cloud Services, which host HALO, HALO AI, HALO Link and HALO AP on Amazon Web Services in deployments the company says are validated by AWS. The cloud page says the customer maintains control over its AWS account while Indica's AWS certified team designs, monitors (with Amazon CloudWatch) and maintains the infrastructure, which places the data in the customer's own tenancy rather than a shared vendor environment. Named cloud deployments include the National Cancer Institute, the University of Washington and the University of Bern's Institute of Tissue Medicine, which the company says was Switzerland's first public health institute to run diagnostic digital pathology in the cloud on AWS.

What is not published is which regions are supported or how the choice is made, and what access Indica's managing team has to the account and from where, which is the processing that could cross a border.

Options stated, with residency implied by the customer's account and the management path left open.

No pricing is published for the platform or for algorithms accessed through the store. Total cost includes cleared scanner hardware, the platform and separately licensed algorithms, and because the clearance is scanner specific, hardware choice and software cost are coupled decisions rather than independent ones.

A broad cleared viewing scope, with the AI's validated scope unpublished. HALO AP Dx is cleared for primary diagnosis in anatomic pathology with named scanners (Hamamatsu NanoZoomer S360MD, Leica Aperio GT 450 DX) and displays, rather than for one tumor type, and HALO AP is CE-IVDR marked for diagnostic use in Europe, the UK and Switzerland. The company serves clinical laboratories, research institutes, contract research organizations and companion diagnostic development, with named clinical users including the National Cancer Institute, Nottingham University Hospitals and the University of Bern. The AI products name their scope (Breast IHC AI for HER2, ER, PR and Ki67 in invasive breast cancer; Lung PD-L1 AI in non-small cell lung cancer; Prostate AI for detection and Gleason grading in core needle biopsies) and list compatible scanner formats, but no validation population or site is published for them on the product pages.

Coverage named with validation behind the cleared viewing scope.