Structure the claim
The claim is normalized into a machine-readable proposition without changing its meaning.
UnmAI
A claim goes in. Evidence, provenance and an auditable verdict come back.
Investigating this claim…
Checking available evidence and provenance. No conclusion is shown until the investigation completes.
Evidence first. The interface displays the backend verdict; it never invents one.
Abstention is a result. Insufficient evidence stays unresolved, never “false.”
Inspectable by design. Evidence and provenance remain attached to every completed investigation.
Method
UnmAI separates evidence gathering from the final decision so the interface can show what supports, contradicts or fails to resolve a claim.
The claim is normalized into a machine-readable proposition without changing its meaning.
Relevant sources are retrieved, ranked and classified by how they bear on the proposition.
Language models may help interpret evidence, but they do not assign the final verdict. The deterministic Verdict Engine remains authoritative.