Smoke runs and ablations prove paths, not real-world claims.
Research and field notes
Methods, artifacts, limitations attached.
This is the working record around FairMind’s research and product thesis—not a pile of generic responsible-AI links. Each item states what kind of evidence exists and what remains unproven.
A useful transformation keeps the limitation visible.
Framework relationships require accountable human review.
Current programs and contracts.
These are bounded descriptions of active work. They are not publication, certification, or production-readiness claims.
Measurement path validated
FairMind-E environmental evidence
A measurement and governance path for energy, carbon, water, uncertainty, mitigation, exceptions, review, and export. The path is validated; real hardware evidence is still required for publication claims.
Read the bounded status
metadata_only
AI BOM fairness evidence
A reviewer-facing transformation of supplied inventory, bias, remediation, evidence, and unknown metadata. It is a metadata profile, not an independent evaluation.
Inspect the source
Recovery target
Evidence Passport
The target portable unit for subject identity, evaluator and metric versions, threshold, raw artifact integrity, result, limitation, reviewer decision, remediation, and superseding rerun.
Inspect the contract
Design principle
Human-reviewed framework reuse
Accepted evidence may support several versioned frameworks. A technical result never accepts its own control mapping or determines conformity.
Compare the roles
A negative result is still useful evidence.
FairMind research should preserve missing measurements, failed hypotheses, incompatible comparisons, provider dependencies, and insufficient data. Deleting those boundaries makes the result less scientific and the product less trustworthy.
Claim strength follows evidence strength.
Have an evaluator, dataset, or assurance question?