Risk-Stratified Pharmacy Governance Framework Proposal for AI-Powered Clinical Systems

While clinical artificial intelligence systems are evolving from chat assistants into tools capable of generating medication recommendations, experts are drawing attention to the risks of autonomous prescription writing. An editorial published in Cureus argues that artificial intelligence must be locally validated and operated under pharmacist supervision.

The Evolution of Artificial Intelligence in Clinical Fields

Clinical artificial intelligence systems are rapidly advancing from chat-based assistance to advanced tools capable of generating medication recommendations and structured clinical actions.

Systems such as AMIE and MIRA, along with tests like RxQA, demonstrate artificial intelligence's simulation success and competence in medication decisions.

Safety Risks of Autonomous Prescribing

Strong performance in simulation tests does not prove that autonomous prescribing is completely safe in the real world and raises serious concerns.

Medication use relies on many factors such as patient-specific conditions, local formularies, drug availability, monitoring capacity, and professional liability.

Proposal for a Risk-Stratified Pharmacy Governance

The proposed risk-stratified management framework mandates that human oversight be increased according to the potential severity and irreversibility of the action.

While low-risk functions can support the process, critical operations such as initiating treatment and adjusting doses must require pharmacist and physician approval.

High-Risk Treatment Processes and Precautions

High-risk treatments such as antimicrobials, anticoagulants, insulin, opioids, and chemotherapy must not be independently altered by artificial intelligence.

Institutions must protect against automation bias and guarantee transparent documentation and traceability.