Enterprise-grade verification infrastructure for AI systems
Ancestor is a deterministic verification engine for AI systems. It evaluates whether a given source supports, contradicts, or fails to justify a claim, producing an auditable trust score with a full reasoning breakdown.
Ancestor accepts three inputs:
A natural-language assertion (e.g. "revenue grew 5%")
A URL or raw document text
Publication date, domain authority
Ancestor computes four weighted signals:
Does the source logically support the claim?
Is the publisher authoritative?
Is the information current?
Is the content structurally credible?
(THE GLASS BOX)
We do not use black box AI scoring. The formula is deterministic:
Score = (entailment × 0.40) + (provenance × 0.30) + (recency × 0.20) + (integrity × 0.10)If Contradiction > 0.5, the Final Score is forced to 0, regardless of how prestigious the source is.
A HIGH-AUTHORITY LIE IS STILL A LIE
(MACHINE IDENTITY ONLY)
anc_live_...Authorization: Bearer anc_live_55a...(ASYNCHRONOUS)
Because deep NLI verification is computationally intensive, the API uses an asynchronous Job pattern.
Response (Immediate):
(THE RESULT)
Once status: "COMPLETED", the response contains:
400Bad Request: Malformed JSON or missing fields401Unauthorized: Invalid/Revoked Key402Payment Required: Billing Past Due or Quota Exceeded413Payload Too Large: Content exceeds 5MB or 20k chars429Too Many Requests: Rate limit hitA Trust Score of 0 is NOT an error.
It is a successful detection of a contradiction.
Use the async job_id to show a "Verifying..." state.
Treat scores between 30–70 as "Human Review Required."
Use the dashboard to rotate keys monthly.
If the API returns 500 or 429, fall back to "Verification Unavailable," not "False."
ANCESTOR PROVIDES SIGNAL, NOT BELIEF