Clinical Accuracy and Safety Concerns Following GPT-5 Public Demonstration in Cancer Care
OpenAI's GPT-5 demonstration showed a patient uploading pathology reports to guide treatment decisions, though privacy implications were not addressed. We evaluated GPT-5 against 100 gastrointestinal oncology cases with tumor-board validation and found identical 85% concordance to GPT-4o, contradicting superiority claims. We recommend mandatory accuracy disclosures and regulatory oversight for AI health demonstrations to protect patient safety and privacy.
Authors
- Michael Bitzer (ORCID: https://orcid.org/0000-0002-4463-8263)
- Carsten Eickhoff (ORCID: https://orcid.org/0000-0001-9895-4061)
- Ivan Capobianco (ORCID: https://orcid.org/0000-0003-3927-591X)
- Derna Stifini
- André L. Mihaljević (ORCID: https://orcid.org/0000-0003-1902-657X)
- Andrea Della Penna
Institutions
- STZ eyetrial (DE)
- University of Tübingen (DE)
Publication Details
- Journal
- Journal of Medical Systems
- Published
- 2025-11-29
- DOI
- https://doi.org/10.1007/s10916-025-02312-x
- Citations
- 1
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 0.65