ESSC-AI Note 1: Evaluation Circuits and Harness-Dependent Capability Reporting

This record contains ESSC-AI Note 1, a minimal relation-structured audit proposal for frontier model evaluation reports. The note argues that an observed model capability should not be interpreted as a property of the model alone, but as an observation produced under an evaluation circuit: the tested system, harness, elicitation method, tool access, resource budget, scoring boundary, and validity checks.

The package includes a PDF technical note and a draft CSV template for recording evaluation-circuit conditions. The proposal is intended as a conceptual reporting aid for capability and safeguard evaluation reports. It is not a validated AI safety benchmark, not a proof of model capability or safety, not a claim that ESSC has been adopted or endorsed by any AI organization, and not a full physical application of ESSC to AI.