Big Ideas,
Real Impact.

The Problem

Clinical AI is already influencing decisions — ambient scribes, radiology LLMs, copilots in every workflow. Model monitoring platforms watch whether the AI is drifting. EHR audit trails capture the signed record. Nobody captures what happened between the AI output and the signature — what was surfaced, whether the clinician found it admissible, what was decided. That moment currently disappears.

The Solution

CAVAL is the open governance standard that captures the attribution record at the moment AI influences a clinical decision. Vendor-agnostic, FHIR R4 compatible, tamper-evident, and designed to sit underneath any AI tool already deployed.

CAVAL is not a decision support tool. It does not suggest or direct care. It documents what happened at the decision point.

The CAVAL Clinical AI Attribution Schema (CAAS) v0.1.2 is published as an open standard under CC BY 4.0.

The Regulatory Reality

UK — NHS DTAC, ICO, and NHS AI governance are moving toward requiring demonstrable audit trails.

EU — The AI Act's Article 14 requires human oversight for high-risk AI. Full enforcement August 2027.

UAE — The National Policy for AI-Driven Smart Healthcare identifies governance, transparency, and accountability as core pillars. Clinical AI is deploying at national scale across NABIDH and Malaffi. The attribution layer doesn't exist yet.

Australia — The TGA requires continuous monitoring for AI scope creep into medical device territory.

US — HIPAA has no framework for AI-influenced clinical decisions. There is no federal requirement to document what AI surfaced, no standard for capturing the clinician's response, and no audit trail requirement at the encounter level. The Trump administration's March 2026 National Policy Framework asked Congress to establish a single federal approach — but until that framework exists, the gap is structural. A health law attorney advising a hospital on HIPAA compliance today cannot point their client to any federal standard for AI attribution.