A Melbourne psychiatrist has implemented a policy requiring new patients to consent to the use of an artificial intelligence scribe to transcribe their therapy sessions, or else seek care elsewhere. The practice has ignited debate over data security, patient consent, and the growing role of AI in healthcare.
AI Note-Taking in Psychiatry
Dr. Hemlata Ranga, a psychiatrist practicing at the Melbourne Clinic in Richmond, has included a clause in her patient registration form stating that she uses AI transcription software, such as Heidi Health AI or Microsoft, to assist with note-taking during appointments and to prepare clinical documents. The form explicitly states that if a patient does not wish to use AI, they must arrange for their referring doctor to refer them to a different service provider.
The use of AI scribes has become increasingly common in the medical field. According to the Royal Australian College of General Practitioners (RACGP), two in five general practitioners now use such tools, and the uptake has doubled in the past 12 months. This trend is partly driven by high demand for mental health services and the administrative burden on clinicians.
Patient Concerns
One patient, who wished to remain anonymous, expressed discomfort with the policy. They noted that the wording of the consent form felt defensive, suggesting that the psychiatrist may have faced objections before and sought to preempt them. “She’s not easily replaceable,” the patient said, highlighting the difficulty of finding a new psychiatrist, especially one recommended by their doctor. The patient has chosen not to seek another provider but remains uneasy about the use of AI in their sessions.
The patient also questioned the accuracy of AI transcriptions, particularly for nuanced therapeutic conversations. They suggested that people should have the option to opt out of such technology without facing barriers to care.
Data Security and Regulatory Gaps
Digital rights experts have raised alarms about the security and privacy of medical data recorded by AI systems. Tom Sulston, head of policy for Digital Rights Watch, pointed out that AI transcription is imperfect and often trained on non-diverse datasets, leading to higher error rates for marginalized groups. He also warned of the risk of data breaches or leaks, which could be particularly damaging for sensitive health information related to sexual or mental health.
“AI is not implemented to improve healthcare outcomes or patient experience, but to reduce administration costs for a clinic,” Sulston said. “In this scenario, patients are not the customer, they’re the product.” He called for stronger regulation to ensure patients have a legally protected right to refuse AI without facing repercussions to their health.
Industry Response
Dr. Tom Kelly, co-founder of Heidi AI, stated that data is processed locally and not used to train the AI or sold to third parties. The company uses third-party testing and auditing to maintain security, but doctors are still responsible for verifying the accuracy of notes. The Melbourne Clinic, where Dr. Ranga practices, confirmed that it does not direct psychiatrists on AI use but expects them to disclose and obtain consent. Psychiatrists at the clinic are independent practitioners.
The RACGP noted that consumer groups have raised concerns about insufficient consent conversations and power imbalances between patients and doctors. Currently, AI scribes are exempt from Therapeutic Goods Administration regulations because they do not diagnose patients, leaving a regulatory gap that experts say needs to be addressed.
