Learning Objectives: 

1. Identify the major failure modes of clinical and generative AI: fabrication, flattery (sycophancy), framing/anchoring effects and degraded attention over long input and recognize how each can compromise diagnostic accuracy and documentation. 

2. Apply practical safeguards on shift: framing prompts neutrally, verifying AI output against a primary source rather than the model, protecting PHI and documenting AI-assisted decisions defensibly when following or overriding an AI alert. 

3. Use the IV-O2-MONITOR-AI framework (Independent Verification, Own the Output, Monitor the Model, Accountable & Indelegable) to maintain clinical and medico-legal accountability for AI-influenced patient care.