CMS issued a Request for Information (RFI) on modernizing CLIA in July 2026, with comments closing September 14, 2026. See the comment search page here:
https://www.regulations.gov/document/CMS-2026-2345-0001
https://www.regulations.gov/document/CMS-2026-2345-0001/comment
>>> Article below by Chat GPT.
##
Pathology Groups Tell CMS How CLIA Should Enter the Digital Age
Summary— CMS and CDC’s 2026 request for information on modernizing CLIA produced substantial agreement among five pathology and laboratory organizations. All treated software, AI, bioinformatics, and clinically consequential data analysis as part of modern laboratory testing, while urging flexible, risk-based oversight rather than rigid technology-specific rules. Important differences remained: whether data-only entities need their own CLIA certificates, whether molecular pathology deserves a new specialty, how explicitly CLIA should require clinical validity, and whether physician review must remain mandatory as AI moves from assistance toward autonomous interpretation.
CMS and CDC Open a Broad CLIA Inquiry
On July 16, 2026, the Centers for Medicare & Medicaid Services and Centers for Disease Control and Prevention published a request for information on possible modernization of the Clinical Laboratory Improvement Amendments regulations. The agencies noted that laboratory technology has advanced substantially since the regulations were implemented in 1992.
The RFI was not itself a proposed rule. Rather, CMS and CDC solicited information that could inform future regulations or guidance. Questions covered pathology-block retention, specimen preparation, suboptimal specimens, test validation, artificial intelligence, data-only facilities, remote competency assessment, emergency preparedness, cybersecurity, and laboratory specialties.
Five organizations—the Association for Molecular Pathology, College of American Pathologists, Coalition for 21st Century Medicine, California Clinical Laboratory Association, and Digital Pathology Association—submitted particularly relevant responses.
A Strong Common Theme: The Test Does Not End at the Wet Bench
Across the five letters, the clearest common position was that modern laboratory testing cannot be divided neatly between physical specimen processing and subsequent computational work. Bioinformatics, image analysis, risk calculations, AI interpretation, and report generation may be integral parts of the test itself.
The organizations generally rejected the idea that analysis becomes something other than laboratory testing merely because it occurs after a slide is scanned, sequencing is completed, or data are transferred to another computer or organization. When software converts specimen-derived information into a clinically actionable result, it remains part of the total testing process.
They also agreed that CLIA modernization should preserve flexibility. Detailed requirements written around today’s technologies could quickly become obsolete. Most favored a framework built around clinical risk, intended use, laboratory-director accountability, end-to-end validation, and professional standards that can evolve more rapidly than federal regulations.
Association for Molecular Pathology:
Bring the Entire Diagnostic Procedure Under CLIA
AMP presented the broadest argument for strengthening CLIA as the principal federal framework for laboratory-developed procedures. It urged CMS to recognize expressly that CLIA is responsible for the quality of all laboratory-developed procedures and to require laboratories to establish their clinical validity.
AMP proposed that clinical validity could be supported through multiple forms of evidence, including peer-reviewed literature, clinical guidelines, case-control studies, registries, real-world data, postmarket evidence, and clinical trials. This would codify an activity that AMP believes responsible laboratory professionals already perform, without importing a single FDA-style approval pathway.
For software and AI, AMP argued that algorithms, machine learning, mathematical formulas, clinical calculations, risk models, and decision-support tools fall within CLIA when they generate, interpret, or materially affect a reportable laboratory result. AMP specifically included SaMS—Software as a Medical Service—within this continuous diagnostic-testing process.
AMP took the clearest position on data-only facilities: a facility receiving specimen-derived data for analysis and generating a reportable clinical result should itself be CLIA-certified. It recommended expanding the regulatory definition of laboratory to cover facilities that examine data, produce interpretations, or summarize information derived from human specimens.
AMP nevertheless opposed highly prescriptive rules. It recommended continued laboratory discretion in validating tests, modifying FDA-authorized assays, preparing reagents, and deciding whether clinically necessary testing can be performed on suboptimal specimens. It also called for new CLIA specialties or subspecialties for molecular, genetic, genomic, and flow-cytometric testing.
College of American Pathologists:
Preserve Pathologist Leadership and Technology-Neutral Rules
CAP’s 25-page response was the most comprehensive. CAP characterized CLIA as an adequate baseline and emphasized the role of accreditation requirements, professional guidelines, and laboratory-director judgment in keeping practice current when federal regulations lag behind technology.
CAP opposed writing rules around specific technologies. In its view, CLIA should regulate quality and accountability through existing specialties, while organizations such as CAP update detailed accreditation checklists and practice guidelines as science changes. Consistent with that philosophy, CAP opposed creating additional CLIA specialties, including a separate molecular pathology specialty.
On artificial intelligence, CAP drew a firm boundary. AI may assist and augment pathologists, identify regions of interest, quantify biomarkers, estimate tumor cellularity, detect mitoses or metastases, integrate molecular and morphologic findings, and generate prognostic or predictive assessments. However, CAP stated that AI systems cannot provide the medical diagnosis or assume responsibility for patient-care decisions. Physician review and approval remain necessary: AI can make predictions, but pathologists make diagnoses.
CAP also placed data-only facilities within CLIA when they generate information used in a final clinical laboratory result. Such activities can include sequence alignment, variant calling, annotation, image analysis, or other processing that directly affects patient-specific findings. Purely administrative, storage, transmission, or other ancillary functions may fall outside the testing process.
Elsewhere, CAP supported considering a longer pathology-block retention requirement but opposed anything beyond ten years, its own accreditation standard. It favored continued exemption of tissue processing and slide preparation from CLIA personnel regulation; laboratory-director discretion for suboptimal specimens; remote competency observation under specified safeguards; and alternative quality procedures for factory-calibrated instruments that users cannot recalibrate.
CAP urged CMS not to create laboratory-specific emergency, biosafety, or cybersecurity mandates where responsibility is shared across hospitals, health systems, vendors, and other healthcare entities. It similarly argued that blood-culture contamination is primarily a preanalytic collection problem that laboratories can monitor but often cannot directly control.
Coalition for 21st Century Medicine:
Risk-Based Reform for Advanced Diagnostics
C21 described molecular testing, distributed testing, dry laboratories, and remote test performance as major gaps in the existing regulations. It supported reasonable, risk-based reform while keeping CLIA the principal quality framework for advanced diagnostic services.
C21 supported longer retention of clinically valuable pathology blocks, noting that many laboratories already retain blocks for ten years or more. It nevertheless recommended considering the specimen type, diagnosis, intended use, and probability of future testing rather than imposing one undifferentiated requirement.
For NGS and other molecular tests, C21 identified performance measures such as limit of detection, variant-class accuracy, depth and uniformity of coverage, precision, specificity, reportable range, and bioinformatics-pipeline performance. It recommended test-specific standards informed by CAP, AMP, ACMG, and similar organizations rather than universal requirements.
C21 viewed algorithms as longstanding components of laboratory testing, not an entirely new phenomenon. Laboratories already use them to analyze digital slides, identify variants, process gene-expression data, and produce proprietary risk scores. It therefore saw no need for a separate AI-specific regulatory regime.
Unlike CAP, C21 cautioned against automatically requiring human review of every software-generated result. Some AI outputs cannot be reproduced or verified visually by a human—for example, an H&E-based prediction of recurrence or molecular status. Their reliability must instead be established through appropriately designed validation studies.
C21 supported bringing clinically actionable data-only work within CLIA. However, it proposed a carve-out for facilities providing only ancillary information that a CLIA-certified laboratory subsequently incorporates, validates, and assumes responsibility for in the final report. It also favored a new, tiered molecular pathology specialty reflecting the large difference between a single-gene PCR assay and a complex NGS test with bioinformatics and AI-assisted interpretation.
California Clinical Laboratory Association:
A Focused Version of the Advanced-Laboratory Position
CCLA’s comments closely paralleled several C21 positions. It emphasized laboratory-director discretion, flexible validation requirements, and recognition that computational interpretation is often inseparable from the laboratory service.
CCLA opposed universal NGS requirements and FDA-style clinical-validity demands that smaller laboratories could not realistically meet. It suggested that CAP and New York State standards could supply practical models without freezing rapidly changing technical details into federal regulation.
CCLA also opposed a separate AI regulatory system. Software used to interpret sequencing data, digital pathology images, or gene-expression results should be validated as part of the complete end-to-end laboratory workflow.
For data-only facilities, CCLA drew the regulatory line at clinical actionability. A facility producing a prognostic, diagnostic, or treatment-related result from an H&E image or sequencing file should fall within CLIA. A facility supplying ancillary information to a CLIA laboratory that validates the input and accepts responsibility for the final result need not necessarily be treated the same way.
Like C21 and AMP, CCLA supported creation of a molecular pathology specialty, potentially divided into tiers based on methodology and complexity. It also called for more consistent national public-health reporting requirements following the fragmented experience of COVID-19.
Digital Pathology Association:
Regulate the Function, Risk, and Degree of Autonomy
DPA concentrated on digital pathology, AI, data-only services, and remote competency assessment. Its central argument was that software is increasingly the mechanism by which a scanned slide becomes a clinically meaningful result.
DPA described three distinct AI roles:
A sole interpreter, producing a result with no human-perceptible equivalent, such as a multimodal recurrence score or inference of molecular status from H&E morphology.
A real-time assistant, directing attention, displaying heat maps, quantifying biomarkers, or triaging cases while the pathologist reviews them.
An independent second reader, comparing its conclusion with the pathologist’s initial interpretation and flagging disagreements.
DPA argued that oversight should depend on which role software actually performs—not merely on the algorithm’s theoretical capabilities. It proposed a tiered system in which requirements increase with clinical consequence and interpretive autonomy.
It also distinguished locked algorithms from adaptive or continuously learning models. Locked products can generally be managed through initial validation and revalidation after material changes. Adaptive systems require ongoing drift monitoring, version control, audit trails, and clear revalidation triggers.
On data-only services, DPA emphasized function rather than physical location or corporate identity. Digitizing a slide is another transformation in the testing continuum, comparable to fixation, sectioning, or staining. An algorithm performs the same clinical function whether it runs inside the originating laboratory or on an outside platform.
DPA stopped short, however, of saying that every organization touching laboratory data should obtain a separate CLIA certificate. It favored having quality expectations follow the interpretive activity across the total testing process, while avoiding indiscriminate expansion of site-based certification.
The Important Areas of Disagreement
The letters revealed four meaningful policy divisions.
First, AMP favored CLIA certification for data-only facilities generating reportable results. C21 and CCLA focused on whether the output is independently actionable, with an ancillary-services exception. DPA focused more heavily on the regulated function than on separate certification of every site. CAP said clinically consequential data-only work should be part of the CLIA-regulated test system but did not fully resolve the certificate mechanics.
Second, AMP, C21, and CCLA supported new molecular or genomic specialties. CAP argued that existing specialties remain adequate and that creating technology-specific categories would make CLIA less adaptable.
Third, CAP insisted on physician review and approval of AI-generated information used in diagnosis. C21 warned against permanently mandating human intervention, while DPA explicitly contemplated systems functioning as sole interpreters or eventually as autonomous diagnostic tools.
Fourth, AMP recommended an explicit CLIA clinical-validity requirement. C21 and CCLA accepted the importance of clinical validity but cautioned strongly against requirements resembling FDA product-approval studies.
Bottom Line
Taken together, the five letters largely rejected a sharp boundary between the physical laboratory and the computational interpretation of laboratory-derived data. They treated digital pathology, bioinformatics, AI, and clinically meaningful data analysis as components of the total testing process—not as unrelated services that begin after laboratory testing ends.
The remaining debate is therefore less about whether these functions require quality oversight than about how CLIA should attach that oversight: to a certified facility, to the laboratory issuing the final report, to the clinical function wherever it occurs, or to some combination of the three. That question is likely to become central to the next generation of CLIA policy.
Source: Public comments submitted in docket CMS-2026-2345, Request for Information; Clinical Laboratory Improvement Amendments of 1988 Regulations.
