Nurses Use ChatGPT and YouTube for Medical Procedures Amid Scary Conditions

Nurses at Royal Darwin Hospital in Australia’s Northern Territory have reported using the AI chatbot ChatGPT to teach themselves medical procedures and calculate medication doses, citing concerns about insufficient training and overwhelming patient loads. The claims, made during interviews with the Australian Broadcasting Corporation, highlight growing reliance on generative artificial intelligence tools in clinical settings despite the absence of formal hospital endorsement or validation for such use.

According to the nurses who spoke with the ABC, they have turned to publicly available AI models like ChatGPT to fill gaps in clinical knowledge, particularly when faced with complex or infrequently performed procedures. One nurse described the working conditions as “scary,” explaining that staff are often left to self-educate using online resources, including YouTube tutorials and AI-generated guidance, due to inadequate support and training protocols.

The use of ChatGPT for medication dosage calculations raises significant patient safety concerns, as the model is not designed or certified for clinical decision-making. While ChatGPT can generate plausible-sounding medical information based on its training data, it lacks real-time access to patient-specific variables, up-to-date pharmacological databases, or regulatory safeguards required for accurate dosing in healthcare environments.

Healthcare professionals and medical AI experts warn that relying on unvetted AI tools for clinical tasks introduces risks of hallucination, outdated information, or incorrect interpretations — especially in high-stakes scenarios like drug administration. Unlike approved clinical decision support systems, which are integrated into electronic health records and subject to rigorous validation, general-purpose AI models operate without oversight in medical contexts.

The situation at Royal Darwin Hospital reflects broader challenges in regional and under-resourced healthcare facilities, where staff shortages and high patient acuity can strain existing training and support systems. Nurses have indicated that escalating workloads and limited access to continuing education contribute to their reliance on informal, self-directed learning methods.

In response to the allegations, NT Health, the government agency responsible for public health services in the Northern Territory, has not confirmed whether any formal investigation into the use of AI for clinical tasks is underway. However, health authorities have previously emphasized that all clinical decisions must be based on approved protocols, peer-reviewed guidelines, and verified medical resources — not unverified AI outputs.

The incident underscores the need for clearer institutional policies regarding the use of generative AI in healthcare environments. As AI tools become more accessible, hospitals and health systems face increasing pressure to establish guidelines that distinguish between appropriate uses — such as administrative support or patient education — and prohibited applications involving diagnosis, treatment planning, or medication management.

Experts in medical ethics and AI safety recommend that healthcare institutions implement strict controls on AI use, including staff training on the limitations of large language models, clear prohibitions on clinical decision-making via unapproved tools, and investment in validated, healthcare-specific AI systems that meet regulatory standards such as those set by the Therapeutic Goods Administration or international medical device frameworks.

As of now, there is no indication that Royal Darwin Hospital or NT Health has endorsed or authorized the use of ChatGPT for clinical purposes. The nurses’ statements suggest a workaround born of necessity rather than policy, reflecting a gap between frontline needs and institutional support in resource-constrained settings.