AI recommends. Humans decide.
Automated output is decision support. Consequential actions — anything that could damage a person’s reputation, restrict their access or be treated as a finding against them — require human approval. Maya does not automate accusation.
An automated score is never a verified credential
A risk indication is an assessment of available evidence at a moment in time. It is not proof, not a verification, and not a legal determination. Maya says so on the result itself, including when no authoritative provider confirmation exists.
Evidence, provenance and expiry
Every signal Maya reasons over carries its source and its age. Community volume alone does not establish fraud; corroboration is required, and stale signals expire rather than accumulating into a permanent mark against someone.
Correction and appeal
Anyone affected by a Maya assessment can ask for it to be corrected or reviewed by a person. See support. A system that can be wrong must be answerable when it is.
What Maya will not do
- Publish a public blacklist of people or numbers.
- Present an automated score as a verified credential.
- Take a consequential action against a person without human approval.
- Ask anyone for a PIN, one-time code or password.
- Claim a provider has confirmed something when it has not.
Data minimisation
Maya asks for what a decision requires and no more. The public website collects nothing at all — see the privacy notice.
Capability honesty
Every public capability carries a maturity label, and nothing currently claims to be LIVE. Where a capability depends on a lawful provider integration that does not exist yet, Maya says so rather than implying the capability is available. See Trust by design.