AI medical scribes may reduce paperwork but could introduce errors into patient records, experts warn

Artificial intelligence tools that automatically generate medical notes are becoming increasingly common in healthcare, but experts warn they may also introduce inaccuracies into patient records and raise concerns about privacy and data security.

According to the report Off the Record by Digital Rights Watch, more than 40% of Australian physicians now use AI-powered medical scribes during consultations. These systems listen to conversations between doctors and patients and use large language models similar to ChatGPT or Claude to produce clinical notes.

For healthcare professionals, AI scribes can significantly reduce administrative workload, allowing more time to focus on patients and helping ease physician burnout. Because Australia's Medicare system reimburses doctors for face-to-face consultations rather than documentation, the technology also offers financial and practical benefits.

However, experts caution that these systems do more than simply transcribe conversations—they summarize and interpret them. As a result, AI-generated notes may contain "hallucinations," producing information that was never discussed during the consultation. For example, advice to quit smoking could be incorrectly recorded as a recommendation to avoid house fires.

AI systems may also reflect biases related to race, ethnicity, socioeconomic background, or language. This could increase the risk of inaccurate medical records for patients from culturally and linguistically diverse communities or for those whose first language is not English.

Early research suggested that up to 90% of AI-generated clinical notes required some level of correction. More recent studies indicate that around 20% contain errors serious enough to potentially influence medical diagnoses.

Researchers note that identifying these mistakes can be challenging because clinicians often work under significant time pressure and may be influenced by automation bias—the tendency to trust computer-generated information without sufficient verification.

The authors also raise concerns about data privacy. Many AI scribe providers do not clearly explain how patient information is stored, who can access it, or which AI models process the data. As a result, healthcare professionals themselves may be unable to answer patients' questions about how their personal medical information is handled.

The article argues that Australian regulators should classify AI scribes as medical devices and require them to undergo the same safety assessments as other medical technologies. The authors also call for stronger privacy protections, greater transparency from AI companies, independent testing of AI systems, and clear accountability for protecting patient data.

According to the authors, without stronger oversight and safeguards, growing use of AI in healthcare could undermine public trust in the medical system.

Follow NEWS.am Medicine on Facebook and Twitter