AI Boosts Diagnosis and Medical Scribing, but Accuracy and Privacy Concerns Remain

Recent studies show that large‑language‑model AI systems are approaching, and in some cases surpassing, doctors in diagnosing complex health problems. An April 2026 study found OpenAI’s o1 model achieved 78% accuracy on diagnostic cases published in the New England Journal of Medicine, and earlier work reported that ChatGPT outperformed physicians on similar tasks. While AI can quickly suggest conditions—such as an ear infection for a child or a possible cardiac issue for an adult—clinicians still need to decide on treatment plans.

In Australia, more than 40% of doctors have begun using AI‑powered scribes that listen to consultations and generate medical notes via large language models. These tools reduce paperwork but raise data‑privacy and accuracy worries. Errors in AI‑generated notes range from minor to significant, with recent analyses indicating that about 20% contain mistakes that could affect diagnoses. Hallucinations, bias against certain demographic groups, and limited transparency about who accesses the data further complicate adoption. Organizations such as Digital Rights Watch warn that without stronger regulation, patient records could be mishandled, undermining trust in the healthcare system.