FDA sets a class II path for AI systems that help read Pap-test slides
A final order published 5 November creates a device category for digital cervical cytology imaging systems with AI, letting future versions clear through the 510(k) route instead of the strictest class III review.
The Food and Drug Administration has created a formal device category for the artificial-intelligence systems that help laboratories read cervical-cancer screening slides. In a final order published in the Federal Register on 5 November 2026, FDA classified the "digital cervical cytology slide imaging system with artificial intelligence algorithm" into class II, the agency's middle regulatory tier, subject to special controls [s1].
What the device is
FDA defines the new generic type as a prescription in vitro diagnostic device: an automated system for creating, viewing and managing digital images of slides prepared from Pap test specimens and conventional Pap smears [s1]. The system works by selecting and presenting areas of interest to facilitate interpretation by the reader, and the AI algorithm "is intended to assist the user in rendering a cytology diagnosis" [s1]. In other words, it is a reading aid for a trained cytologist, not a stand-alone diagnostic machine.
The classification is codified as a new regulation at 21 CFR 864.3900, and the order is effective 5 November 2026 [s1]. Because these are prescription in vitro diagnostic products, they remain subject to the prescription labelling requirements for such products [s1].
Why class II matters
Under US device law, any genuinely novel device is automatically placed in class III, the category reserved for the highest-risk products, which generally requires the demanding premarket approval process. FDA can move such a device into a lower class through "De Novo" classification, the route authorised under section 513(f)(2) of the Federal Food, Drug, and Cosmetic Act [s1]. This order is the product of that process: FDA determined that class II with special controls provides a reasonable assurance of the device's safety and effectiveness, and that classification was applicable as of 31 January 2024 [s1].
The practical consequence is forward-looking. Once a device type is classified into class II, later products of the same type can generally reach the market through the 510(k) premarket notification pathway rather than facing automatic class III review [s1]. FDA framed the move as one that will "enhance patients' access to beneficial innovative devices, in part by reducing regulatory burdens" [s1].
The special controls
Class II devices are not lightly regulated; they carry binding special controls that address the risks FDA has identified. The agency named the central risk as false negative and false positive cervical-cytology results, and the order requires measures to mitigate it [s1]. Among the controls, clinical performance testing must show that the device using its AI algorithms is non-inferior to an appropriate comparator method, such as reading by optical microscope, and each clinical study must document relevant details including the origin of the study slides and images, reader and annotator qualifications, and the method of annotation [s1]. The labelling required for these products must specify the intended use, including the specimen type, and carry information on device performance [s1]. A device must comply with the named special controls to fall within the class II classification and avoid automatic placement in class III [s1].
What it means for screening
Cervical screening has been shifting toward digital workflows, in which slides are scanned and read on screen rather than down a microscope, sometimes with software flagging suspect cells. This order does not approve any particular product or endorse a brand; it sets the regulatory category and the evidentiary bar that such systems must meet [s1]. For laboratories and the people whose Pap tests they process, the significance is structural: FDA has decided these AI reading aids belong in the middle risk tier, with performance that must be shown non-inferior to established reading and with labelling that states how the device performs.
What is settled
What the order fixes is concrete. A dated final order now exists, the device type is class II, the classification lives at 21 CFR 864.3900, the special controls are codified, and the order took effect on 5 November 2026 [s1]. What it does not do is clear any specific product for sale or change how a screening result should be acted on clinically. The AI remains, by the device's own definition, an aid to a human reader rendering the diagnosis rather than a replacement for one [s1]. The order's weight lies in setting the category and the evidentiary bar, which every future device of this type must meet before it can be marketed [s1].
Sources
- Medical Devices; Hematology and Pathology Devices; Classification of the Digital Cervical Cytology Slide Imaging System With Artificial Intelligence Algorithm — U.S. Food and Drug Administration / Federal Register , November 5, 2026
FDA classifies software that estimates a patient's odds of progressing to dementia
A final order published 5 November creates a class II category for prognostic software that weighs multiple clinical inputs to gauge the chance of progression from mild cognitive impairment — for adjunctive use only.
FDA creates a class II category for wearable bone-vibration devices
A final order published 5 November classifies wearable devices that apply targeted mechanical vibration to the skeleton to reduce loss of bone strength or density, setting the evidence such products must meet.
FDA creates a class II category for at-home STI test kits read in the lab
A final order published 9 October classifies systems that let people collect their own specimens for non-viral STI testing as class II, moving them out of the automatic class III that demands premarket approval.
FDA writes a device class for gene panels that gauge inherited cancer risk
A final order codifies Class II status for high-throughput DNA sequencing systems that detect germline mutations in cancer genes, built on Invitae's panel. It becomes a predicate other makers can clear against.