FDA Approves First Fully Autonomous AI Pathologist for Cancer Screening

The U.S. Food and Drug Administration has cleared the first fully autonomous AI system for primary diagnosis in surgical pathology. The platform, developed by Paige.AI and trained on more than 5 million whole-slide images, can sign out routine cancer-screening cases — including prostate biopsies, breast core biopsies, and skin excisions — without mandatory human review.

What the Clearance Covers

The clearance is specific: the AI can independently report negative findings on routine cases in the approved indications. Positive findings, ambiguous cases, and any tissue outside the cleared indications are escalated to a human pathologist. The system must also generate an audit trail that captures the model's reasoning and confidence scores for every case.

"This is the first time the FDA has been willing to put its name behind a 'no human in the loop' diagnostic clearance. The bar is appropriately high, and Paige has cleared it," said Dr. Jeffrey Golden of Brigham and Women's Hospital.

Why It Matters

The U.S. faces a structural shortage of pathologists, with training pipelines not keeping pace with retirements. Average turnaround times for biopsy results have lengthened to 7-10 days in many systems. The AI platform is expected to cut routine case turnaround to under 24 hours by absorbing the high-volume, low-complexity workload.

• Cleared indications: Prostate, breast, and skin specimens
• Turnaround target: <24 hours for routine cases
• Training set: 5M whole-slide images
• Audit requirement: Confidence score + reasoning chain per case

Reception

Pathology groups have reacted with cautious acceptance. Most agree that high-volume routine screening is the right place to start, and several academic centers have announced pilot integrations. Professional societies are drafting new training standards for pathologists who will spend more of their time on complex cases and AI oversight rather than routine screening.

The clearance is also a watershed for clinical AI more broadly. It signals the FDA's willingness to authorize systems that operate without mandatory human review, provided that the risk profile is favorable and the failure modes are well characterized. Expect similar clearances in radiology and dermatology within the next 18 months.