Submit
An AI-generated or AI-assisted output is submitted with its intended use.
In development · AI assurance
Attested by Humans.
A human review layer for AI-generated and AI-assisted work.
Objective
AI can generate. Humans remain accountable.
Attested by Humans explores a simple question: when AI produces useful work, how do we make meaningful human scrutiny visible?
The initiative focuses on defined review scopes, explicit criteria, traceable issues and reviewer judgment rather than trying to prove whether something was created by AI.
Core principle
Attested does not mean AI-free.
It means a human reviewed a defined output against stated criteria and recorded an assessment.
Not a claim that a human made it. A statement that a human checked it.
The attestation should state what was reviewed, what criteria were applied, what limitations remain, and what judgment was reached. It does not imply authorship, infallibility or blanket certification beyond that scope.
Review dimensions
What a human review may examine.
The framework is intended to stay explicit and inspectable rather than collapse everything into a single opaque score.
Accuracy
Are factual claims correct and appropriately qualified?
Evidence
Are claims supported by credible and traceable sources?
Reasoning
Do assumptions, logic and conclusions hold together?
Completeness
Are material gaps or missing perspectives visible?
Transparency
Are uncertainty, limitations and AI involvement clear?
Fitness for purpose
Is the output suitable for the decision or use intended?
Concept flow
From output to accountable review.
The current direction is deliberately simple.
Review
A human examines the output against a declared set of criteria.
Record
Scores where useful, issues, uncertainty and recommendations are documented.
Attest
The result becomes a visible record of what was checked and by whom.
Direction
A human trust layer for an AI-made world.
Initial work focuses on review methods, attestation formats and early pilots. Later versions may support persistent attestation IDs, reviewer identity, machine-verifiable records and domain-specific review methods.