What the tool can do
- Show whether each minimum evidence field is populated
- Keep source, evaluation, artefacts and boundary in one record
- Import, copy and download handoff-ready browser-local JSON
FUURAA AI Knowledge Library · Browser-local tool
AI Claim Evidence Record Builder
Turn a public AI claim into a handoff-ready, reviewable and bounded structured record. Entries remain in this browser page unless you choose to download or copy them.
Record workspace
Seven sections reconnect a conclusion to its source and boundary.
Start blank or load a fully marked fictional example. Completion measures only whether required fields are present; it does not judge factual truth.
0%32 fields remain incomplete.
Claim and decision
Write one bounded claim and name the decision it is allowed to inform.
Subject identity
Freeze the exact system, release and operator under review.
Source record
Preserve the canonical record and the relationship between publisher and subject.
Evaluation design
Make the task, comparison, sample, metrics, uncertainty and failure rules reviewable.
Technical artefacts
Identify the data, model, code and configuration another reviewer would need.
Applicability boundary
State where the evidence applies and which transfers remain untested.
Evidence state and review
Record a dated judgment, its tensions and the trigger for reopening it.
JSON
Machine-readable record preview
The export uses RFC 8259 JSON. The field structure is a FUURAA public method, not an international standard or certification format.
{
"schema": "fuuraa.ai-claim-evidence-record/v1",
"record_id": "",
"claim": {
"statement": "",
"decision": ""
},
"subject": {
"system": "",
"version": "",
"operator": ""
},
"source": {
"canonical_url": "",
"type": "",
"publisher": "",
"published_at": "",
"accessed_at": ""
},
"evaluation": {
"task": "",
"baseline": "",
"sample": "",
"metrics": "",
"uncertainty": "",
"failure_criteria": ""
},
"artefacts": {
"data": "",
"model": "",
"code_and_environment": "",
"configuration": ""
},
"applicability_boundary": {
"population": "",
"language": "",
"geography": "",
"workflow": "",
"excluded_uses": ""
},
"judgement": {
"evidence_state": "",
"rationale": "",
"contradictions_and_unknowns": "",
"review_owner": "",
"reviewed_at": "",
"next_review_trigger": ""
}
}Applicability boundary
A complete record does not mean the claim has been proved.
What the tool cannot prove
- That a source is true, a file untampered or a publisher identity valid
- That study design is valid, results reproducible or a system deployment-ready
- Legal, regulatory, safety, audit or certification compliance
FUURAA analysisThe most reliable record is not the one with the most fields; it is the one whose claim, evidence and decision have matching granularity. Evidence from a fixed English test set may support that test result, but should not become approval for multilingual real-world deployment.
Primary sources and boundaries
The record method connects to public specifications without claiming to implement them normatively.
26 January 2023
NIST AI RMF 1.0
Connects evidence work to context, affected people, measurement, governance and continuing risk management.
BoundaryVoluntary and cross-sector; it does not prescribe this record format or confer certification.
Open primary source ↗26 July 2024 · page updated 8 April 2026NIST AI 600-1 · Generative AI Profile
Adds generative-AI concerns including provenance, measurement, information integrity and human oversight.
BoundaryA companion profile whose actions still require prioritisation for the actual system and use context.
Open primary source ↗W3C Recommendation · 30 April 2013W3C PROV-DM
Provides a conceptual model for entities, activities, agents, derivations and responsibility in provenance records.
BoundaryThis tool borrows provenance concepts; its exported JSON is not a normative PROV serialisation.
Open primary source ↗2019Model Cards for Model Reporting
Motivates disclosure of intended use, evaluation conditions, performance and relevant subgroup results.
BoundaryDocumentation cannot independently verify developer measurements or make an unsuitable system safe.
Open primary source ↗23 March 2018 · version 8 on 1 December 2021Datasheets for Datasets
Structures questions about dataset motivation, composition, collection, use and limitations.
BoundaryDisclosure quality depends on the record supplied and does not replace rights review or direct data inspection.
Open primary source ↗Internet Standard · December 2017IETF RFC 8259 · JSON
Defines the portable structured-data interchange format used by the exported record.
BoundaryValid JSON syntax does not establish the truth, completeness or authenticity of its contents.
Open primary source ↗Method and evidence atlas