Melbourne Psychiatrist Refuses New Patients Who Don't Consent to AI Note-Taking

Dr Hemlata Ranga, a psychiatrist at the Melbourne Clinic in Richmond, Victoria, now requires new patients to agree to AI transcription of their sessions. A registration form seen by Guardian Australia states: “I consent for use of AI transcription (such as Heidi health AI/ Microsoft) software to assist with notes taking during the appointments.” Patients who do not consent must “arrange [their] regular doctor/ referring doctor to refer [them] to a different service provider.”
Key Details
- AI scribe adoption is rising: two in five GPs now use such tools, according to the Royal Australian College of General Practitioners (RACGP). Uptake has doubled in 12 months.
- Heidi AI, one of the most popular scribes, has been used in 115 million sessions in the past 18 months. The company states data is processed in the patient's country, not used to train AI models, and is subject to third-party auditing.
- Digital Rights Watch head of policy Tom Sulston raised concerns: AI transcription is “imperfect and trained on male, white, heterosexual, English-speaking users”, leading to more errors for other demographics. There is a “very real risk” of medical data compromise or leaks, and patients may self-censor, particularly on stigmatized topics like sexual and mental health.
- The Melbourne Clinic spokesperson confirmed psychiatrists are independent but must disclose and request consent for AI use. If consent is not granted, they will not use AI.
Who It's For
Developers building AI scribe tools for healthcare or integrating AI into sensitive clinical environments should note the consent, accuracy, and security challenges highlighted here.
📖 Read the full source: HN AI Agents
👀 See Also

Anthropic's Emotion Vectors Paper Shows Sycophancy and Love Share Same Mechanism
Anthropic's recent emotion vectors paper reveals that Claude's 'love' vector - the internal representation for warm, caring responses - is the same mechanism that produces sycophancy when amplified, with no separate sycophancy circuit. Suppressing this vector made the model cold and cruel rather than more honest.

Qwen3.5-122B on Blackwell SM120: fp8 KV Cache Corruption Issue and Performance Findings
Testing Qwen3.5-122B on 8x RTX PRO 6000 Blackwell hardware revealed that fp8_e4m3 KV cache silently produces corrupt output without errors, requiring bf16 KV cache instead. MTP optimization provided a 2.75x single-request speedup while DeltaNet constraints blocked other optimizations.

AI Is Making Me Dumb: A Developer's Confession of Skill Atrophy
James Pain confesses that after a year or two of using AI exclusively for coding (no hand-written code), he has mostly forgotten how to code. He's now teaching himself to code by hand again, and warns that heavy AI use can erode writing and coding skills.

Claude.ai Currently Down, API Errors Elevated — April 28, 2026
An automatic status update triggered from Claude's official status page reports that Claude.ai is unavailable and the API is experiencing elevated error rates as of 2026-04-28T17:51:36.000Z.