AI Policy

1Why this page exists

VERA™ operates a public registry about how AI systems represent brands, people, and content. A registry that holds others to account for AI use should state its own. This page describes where AI appears in our systems, how AI-derived information is labeled, and what we never let AI decide.

2Where we use AI

We use AI models for measurement and drafting: sampling how public AI platforms answer questions about registered entities, scoring content signals during verification (for example the probability that media is AI-generated or manipulated), extracting candidate facts from public sources, and drafting newsroom summaries of third-party reporting.

We route this work through our own orchestration layer and paid model access. Model output is treated as an instrument reading, never as ground truth.

3The three-namespace rule

Every field in the registry belongs to one of three classes, and they never mix. Authoritative facts come from verified documents and owner-verified claims; AI never writes them. Measurements are AI-derived readings (visibility, sentiment, detection scores) and are labeled as measurements, with the date they were taken. Verdicts on disputed or ambiguous matters are made by people, recorded with reviewer accountability.

If a value in the registry was produced by a model, it is presented as a measurement. Absence of verification is itself displayed honestly; we do not fill gaps with AI guesses.

4What AI never decides

AI does not decide verification outcomes, registry standing, badge eligibility, editorial publication, or the resolution of disputes. Signed credentials are issued by an automated policy layer that enforces human-set rules and refuses to sign anything that does not read back exactly as authorized; the rules themselves are set and reviewed by people.

5AI-generated content, ours included

Our verification systems classify AI-generated media, and when our own published or credentialed content is AI-generated, that status is disclosed, including inside signed credentials, where the AI-generated flag is part of the signed record. Newsroom items drafted with AI assistance are reviewed by a human editor before publication and attribute every claim to its source publication.

6Your content and our models

We do not sell your data, and we do not use content submitted for verification to train AI models. Content you submit is analyzed to produce your verification result and retained as evidence of that result, as described in the Privacy Policy.

7Provenance is not truth

A VERA verification or credential records what was checked, by whom, when, and with what evidence. It does not certify that a claim is true, and we do not present AI analysis, ours or anyone's, as a verdict on truth. Where our records rely on AI measurement, the confidence and method class are part of the record.

8Questions and corrections

If you believe an AI-derived value in the registry is wrong, contact corrections@veraregistry.com and it will be reviewed by a person. This page changes as our systems change; material changes update the effective date above.