Preserve the three parts of the record.
The initial AI output remains preserved. The Huxpert report presents the expert’s findings, discrepancies and supported corrections. A new output produced after a change is a third item, identified and checked separately.
The expert’s correction does not retrospectively improve the initial output’s Hu-score. A new grade requires sufficient new verification, with its own date, object and scope.
Find the right report version.
The reference or QR code of a registered report opens its minimal public record: version, publishable scope, responsible person and status. Detailed evidence and client documents remain private. Superseded versions remain identifiable.
Comparing the SHA-256 hash checks whether a PDF matches the registered file. The hash alone does not establish the truth of a photograph or finding. The comparison takes place locally in the browser, without uploading the PDF.
Source and scope
Documenting findings and tracking changes draws on the evaluation approach in NIST AI RMF 1.0. The registry statuses, QR code and Hu-score are Huxpert implementation choices, without NIST endorsement. NIST AI RMF 1.0.
Related reading
Have an AI-generated result checked
Identify the output, actual object and checks required before requesting a quote.