GPT-5 proposes oncocytoma first in two retrospective cases

AI system

System: ChatGPT (GPT-5, 2025 version)
Model: GPT-5.
Release: Exact release not established.
Application developer: OpenAI
Role: GPT-5 proposed diagnoses from images and brief clinical descriptions in a retrospective test.
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: Slit-lamp photographs plus a brief clinical summary submitted to a multimodal model.
Human oversight: Not assessed
Relation to AI: GPT-5 produced the correct primary hypothesis in two retrospective cases when given images and minimal clinical data. The result is a proof of concept reported by non-vendor clinicians. It had no role in the patients' management.
AI contribution: GPT-5 produced the correct primary hypothesis in two retrospective cases when given images and minimal clinical data. The result is a proof of concept reported by non-vendor clinicians. It had no role in the patients' management.

Outcome identifiers: Not assessed
Benefit: Sassari ophthalmologists gave ChatGPT (GPT-5) slit-lamp photos and brief summaries of two surgically confirmed caruncular oncocytomas; it proposed oncocytoma first in both. Retrospective proof only.
Category: Improved Task Accuracy
Status: Not assessed
Beneficiary groups: People. Organizations. Society.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Participants: Developer: OpenAI. Operator: Department of Medicine, Surgery and Pharmacy, University of Sassari.
Jurisdictions: Benefit: Sassari, Italy. Developer: United States. Operator: Italy.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Conditions for replication: Prospective testing before any clinical use. Larger samples than two cases.
Replication attempts: Not assessed
Later findings: The report is a retrospective proof of concept on two histologically confirmed cases; the cited source describes no follow-up and the authors call for larger validation and benchmarking against human experts.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Assessed benefits: People. Organizations. Society.
Depth: Not assessed
Reasons: People: No effect on the two patients, whose diagnoses were already confirmed. Organizations: No organizational outcome. Society: No attributable event.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: The deployer's account

Kind: Not assessed
Related entries: Not assessed
Field: Health; ophthalmology. Patients with rare ocular adnexal tumors (retrospective).
Purpose: Test whether a multimodal large language model can generate diagnostic hypotheses from slit-lamp images
Dates: Date basis unassessed: 2025 (precision unassessed). Updated: 21 October 2026.
Stage: Not assessed
Version: SHA-256: e45032abd8e2…