AI Transparency Guidelines for FHIR Implementation

This document is a project proposal focused on AI transparency within the FHIR framework, specifically addressing the requirements set forth by the HHS EHR Reporting provision of the Century Cures Act. The proposal outlines the need for guidance on representing AI-generated content in FHIR resources, ensuring consistency for downstream system utilization. Key elements include defining AI inferences, model identification, versioning, generation timestamps, confidence scores, human review status, and model limitations. Additionally, it proposes the development of a FHIR Implementation Guide to standardize the use of generative AI and large language models (LLMs) in FHIR resources and operations. The document also discusses establishing standard patterns for FHIR operations involving AI, including operation definitions, parameters, input validation, output formatting, error handling, and audit trail requirements. The proposal seeks feedback on its relevance and interest in sponsorship from various work groups.