CANON Named Owner Principle · every AI deployment requires two named persons in the audit trail, a Governance Owner and a Decision Owner, not one substituting for the other WORKING PAPER №01 The handoff that isn’t · how clinical AI escapes accountability · Mo Johnson, MD MBA EVIDENCE Duke-Margolis 2026 · most US health systems have not named who owns the clinical AI decision when something goes wrong CANON Layer 4 · Clinical AI Governance at the bedside · the layer where the named owner has to live FRAMEWORK Clinical AI Accountability Canvas™ · the diagnostic framework distinguishing Clinical AI Governance from General AI Governance EVIDENCE Stanford MedAgentBench · agentic systems already executing clinical recommendations without a named adjudicator on the chart CANON The Two Inputs · internal data the institution audits · external data the model was trained on, rarely audited at the deployment site CITATION MedicoVigilance™ Issue 6 · The Layer With No Name · 1,627 institutional subscribers CANON The Accountability Gap™ · the structural failure point where AI stops and the named owner starts FRAMEWORK Mind the 9 Blocks™ · the nine institutional blocks that must be in place before clinical AI deployment EVIDENCE npj Digital Medicine · the four-layer governance cascade · most institutions have built the first two layers and left Layer 4 unbuilt FRAMEWORK The Accountability Gap™ (TAG™) · two seats, one handoff, any regulated sector · clinical AI and financial AI WORKING PAPER №02 The seam · how a bank's automated triage system closed the alerts nobody read · Mo Johnson, MD MBA FRAMEWORK The Accountability Canvas · nine blocks, one Gap Score™, any regulated deployment · clinical AI and financial AI PORTFOLIO GPe Research · adjudication infrastructure for regulated sectors, in clinical AI and financial AI. CANON Named Owner Principle · every AI deployment requires two named persons in the audit trail, a Governance Owner and a Decision Owner, not one substituting for the other WORKING PAPER №01 The handoff that isn’t · how clinical AI escapes accountability · Mo Johnson, MD MBA EVIDENCE Duke-Margolis 2026 · most US health systems have not named who owns the clinical AI decision when something goes wrong CANON Layer 4 · Clinical AI Governance at the bedside · the layer where the named owner has to live FRAMEWORK Clinical AI Accountability Canvas™ · the diagnostic framework distinguishing Clinical AI Governance from General AI Governance EVIDENCE Stanford MedAgentBench · agentic systems already executing clinical recommendations without a named adjudicator on the chart CANON The Two Inputs · internal data the institution audits · external data the model was trained on, rarely audited at the deployment site CITATION MedicoVigilance™ Issue 6 · The Layer With No Name · 1,627 institutional subscribers CANON The Accountability Gap™ · the structural failure point where AI stops and the named owner starts FRAMEWORK Mind the 9 Blocks™ · the nine institutional blocks that must be in place before clinical AI deployment EVIDENCE npj Digital Medicine · the four-layer governance cascade · most institutions have built the first two layers and left Layer 4 unbuilt FRAMEWORK The Accountability Gap™ (TAG™) · two seats, one handoff, any regulated sector · clinical AI and financial AI WORKING PAPER №02 The seam · how a bank's automated triage system closed the alerts nobody read · Mo Johnson, MD MBA FRAMEWORK The Accountability Canvas · nine blocks, one Gap Score™, any regulated deployment · clinical AI and financial AI PORTFOLIO GPe Research · adjudication infrastructure for regulated sectors, in clinical AI and financial AI.

TOPICS

The territory of accountability in regulated AI.

Eight subject areas define what GPe Research Publications takes up and what it leaves to other venues.

Editorial taxonomy

The publication works inside a specific territory. The questions it takes up are structural, not technical. The boundary is decision ownership. Anything upstream of the decision belongs to model developers, regulators, and the informatics teams of the deploying sector. Anything downstream belongs to liability systems and the courts. GPe Research Publications works the layer in between, where the AI stops and the institution has to stand behind the output.

The eight subject areas below organize that layer. Each defines a recurring locus of failure or a recurring instrument for closing it. Papers are tagged accordingly and accumulate over time into the institutional record of the field.

Subject areas

Accountability Architecture

The structural design of who owns what when an AI system produces an output. Named owners, audit trails, escalation paths, decision traceability.
Clinical AI Governance

The institutional layer above deployment. Board oversight, committee structure, policy frameworks, vendor management, procurement standards.
Financial AI Governance

The accountability architecture for automated decision and monitoring systems in financial institutions, naming the Governance Owner and Decision Owner across triage, escalation, and review workflows.
Diagnostic Frameworks

The instruments the publication develops to name where accountability breaks. The Clinical AI Accountability Canvas™. Mind the 9 Blocks™. MedicoVigilance™. FinVigilance™.
Medical Legal

Where clinical accountability meets legal and financial exposure. What the record shows when a decision is examined after the fact: what an insurer will cover, what a regulator will ask for, and what a clinician can defend.
Named Owner Principle

The editorial spine of the publication. The case for mandatory professional ownership of every AI generated recommendation, above the clinician in the audit trail.
Audit Trail Integrity

Decision traceability at the point of decision. What gets logged, what gets named, what survives discovery, what holds up in litigation.
Regulatory Standing

The interface between regulated AI deployment and federal regulators, sector specific licensing and accreditation bodies, and tort law. Where institutional liability lives.

What this publication does not cover

The publication does not run model performance benchmarks. It does not evaluate AI products. It does not adjudicate vendor claims. It does not publish opinion that is not grounded in clinical, financial, legal, or institutional standing. The work upstream of the decision belongs to model developers and the informatics journals of the deploying sector. The work downstream of harm belongs to liability law and the safety literature of the deploying sector. The publication holds the layer in between.