The Agentic Accountability Baseline
Issued: July 22, 2026Status: ACTIVE
AI Governance
+12
Coca-Cola disclosed a ransomware attack in Item 8.01. The rule built for this has a number: 1.05
+11
AI-Assisted Research, No Human Oversight, No Audit Trail
A fraudulent model reached 244,000 downloads. The controls returned "verified." The model was not authorized. This report maps the gap between those two facts.
+10
244,000 Downloads. Zero Verification.
The Obligation Flows One Way. So Does the Data.
The standard exists. The instruments are published. The scoring has begun.
The harness captures what the agent did.The harness captures what the agent did. This defines what sufficient looks like.
+9