Security and customer control

Verified Identity is designed so automation removes repetitive work without removing customer authority.

Account protection

Google provides authentication. Customer data access is checked again in the server-side data layer and database row policies using organisation membership.

Approval boundary

AI may diagnose and draft, but cannot approve its own work. Only an owner or approver can decide. Every decision is bound to one immutable version; changing the wording requires a new decision.

Controlled execution

Approved work enters an idempotent, leased queue. Publishing is limited to configured HTTPS destinations on a server allowlist. Secrets stay in the host environment, not the customer browser or database destination record.

Evidence and recovery

Successful connectors must return before and after snapshots plus a rollback reference. Execution, failure, actor, destination, and remeasurement events are recorded for review.

Measurement privacy

External AI measurement is off by default for every organisation. An owner or approver must enable the named provider after seeing the public-data scope. Answer generation is kept separate from target classification, and future transfers stop when measurement is disabled.

Platform separation

Recommendation information cannot alter Atlas Eligibility or Verified Dispatch. Verified Delivery observations are quarantined for review and cannot automatically change trust or verification.

Report a concern

Report suspected unauthorised access, unsafe publishing, or data exposure to info@stripoutlondon.co.uk. Do not include passwords or private keys in email.