Normal view

Received — 6 April 2026 ⏭ npj Digital Medicine

npj Digital Medicine
New model, old risks: sociodemographic bias and adversarial hallucinations vulnerability in GPT-5 Mahmud Omar · Reem Agbareia · Donald U. Apakama · Carol R. Horowitz · Robert Freeman · Alexander W. Charney · Girish N. Nadkarni · Eyal Klang
npj Digital Medicine, Published online: 04 April 2026; doi:10.1038/s41746-026-02584-8We re-evaluated GPT-5 using our published pipelines: 500 emergency vignettes across 32 sociodemographic labels for bias, and adversarial prompts with fabricated details. GPT-5 showed no measurable improvement over GPT-4o in sociodemographic-linked decision variation, with several LGBTQIA+ groups flagged for mental-health screening in 100% of cases. Adversarial hallucination rates were higher (65% vs 53% for GPT-
4 April 2026 at 08:00

New model, old risks: sociodemographic bias and adversarial hallucinations vulnerability in GPT-5

By: Mahmud Omar · Reem Agbareia · Donald U. Apakama · Carol R. Horowitz · Robert Freeman · Alexander W. Charney · Girish N. Nadkarni · Eyal Klang

4 April 2026 at 08:00

npj Digital Medicine, Published online: 04 April 2026; doi:10.1038/s41746-026-02584-8

We re-evaluated GPT-5 using our published pipelines: 500 emergency vignettes across 32 sociodemographic labels for bias, and adversarial prompts with fabricated details. GPT-5 showed no measurable improvement over GPT-4o in sociodemographic-linked decision variation, with several LGBTQIA+ groups flagged for mental-health screening in 100% of cases. Adversarial hallucination rates were higher (65% vs 53% for GPT-4o); a mitigation prompt reduced this to 7.67%.