Safe, Ethical, Certified, Uncertainty-Aware, Robust, and Explainable AI for Health

This book constitutes the refereed post-conference proceedings of the First International Workshop, on Safe, Ethical, Certified, Uncertainty-aware, Robust, and Explainable AI for Health, SECURE AI4H@AAAI 2026, Held in Conjunction with AAAI 2026, held in Singapore, during January 27, 2026.

The 16 full papers and 3 short papers included in this volume were carefully reviewed and selected from 60 submissions. The papers cover the following topical sections: Uncertainty, Calibration, and Reliable Prediction; Robustness, Fairness, and Explainability in Clinical Models; Trustworthy Large Language Models and Agents in Health; Governance, Security, and Deployment.

Inhalt

.- Uncertainty, Calibration, and Reliable Prediction.
.- Uncertainty-Aware Reward and Replay for Trustworthy Semi-Supervised MRI Anomaly Detection.
.- Adaptive Conformal Prediction via Bayesian Uncertainty Weighting for Hierarchical Healthcare Data.
.- BLUFF-1000: Measuring Uncertainty Expression in RAG.
.- GUARDIAN: Gated Uncertainty-Aware Runtime Dual Invariants for EEG-Controlled Agents.
.- AI, the SIR Model, and Data Sharpened Kernel Smoothing.
.- Robustness, Fairness, and Explainability in Clinical Models.
.- Towards Robust and Fair Next Visit Diagnosis Prediction under Noisy Clinical Notes with Large Language Models.
.- Responsible AI Evaluation in Clinical Machine Learning: A Study on In-Hospital Mortality Prediction Using MIMIC-IV.
.- DERM-Clu: A Diffusion-based Explainable Representation Model using Clustering for Skin Lesion Classification.
.- Emergent Zero-Shot Global OOD Performance in Multimodal Mammography Models.
.- Robust and Interpretable Multimodal Fusion in Med-VQA: A Perturbation Benchmark for Clinical Safety.
.- When Patches Learn the Domain: Adaptive Vulnerabilities in Medical AI.
.- SPACE: Sparsely Primed Autoencoder Counterfactual Explanations for Single-Cell Gene Perturbation Experiments.
.- Interpretable Multimodal ECG Classification using RAG-Based Cross-Attention Fusion.
.- Trustworthy Large Language Models and Agents in Health.
.- Side Effects May Include: Alignment Decay Post Medical Fine-Tuning.
.- Trust-X: Towards Transparent and Safe Multi-Agent Reasoning
.- Assessing the Quality of Mental Health Support in LLM Responses through Multi-Attribute Human Evaluation.
.- EmoMed: An Emotionally-Aware Agent for Multimodal Medical Support with Real-Time Information Retrieval.
.- Governance, Security, and Deployment.
.- Regulatory Frameworks for AI in Medical Devices: Cross-Jurisdictional Perspectives on Certification, Implementation, and Post-Market Surveillance.
.- FAIR Voice Biomarker Data for Secure and Trustworthy GenAI in Health: A Cross-Domain Assessment.