Security and trust

Guardrails that fail closed and leave receipts.

Personal privacy and regulated workflows are enforced by application permission checks, default-deny grants, approval packets, and receipts; database-level controls protect direct API paths where implemented, but Cuadrila does not claim autonomous regulated decision-making.

Regulated decisions

Cuadrila is software that prepares, organizes, and routes work. Its AI coworkers draft communications, assemble documents, and propose actions with sources attached — but they do not make legal, medical, financial, HR, or compliance decisions, and they do not provide professional advice. Every regulated or externally consequential action is approval-gated and fails closed: a named human reviews the proposed action, its sources, and its receipt before it executes. You remain responsible for deciding whether an output is appropriate for your business, your industry, and your legal obligations.

Backed guardrails

008

Default-deny grants

Coworkers act only through scopes and tools that are granted for the current context.

Evidence: 008 grant permission model

015

Approval packets

Externally consequential actions are parked with a named packet until an authorized human decides.

Evidence: 015 approvals everywhere

009

Durable receipts

Actions leave receipts with source, scope, and decision posture so teams can inspect what happened.

Evidence: 009 artifact receipt entity

Will not do

  • Make legal, medical, financial, HR, or compliance decisions

    003 regulated-decision disclaimer plus 015 approval gates

  • Send, file, publish, or share externally without the required approval state

    015 approval packets and 008 verb-tier grants

  • Treat marketing-only characters as shipped product agents

    003 two-rosters hygiene and 007 active product roster

  • Claim generated media, public apps, local execution, help center, or backup as active without launch evidence

    028 hardening evidence and 029 go/no-go review

Trust destinations

Support

Support destination is configured from the platform brand record.

Open destination

Updates

Operational update host is configured from the platform brand record.

Open destination

Subprocessors

Subprocessor posture is summarized in the privacy policy until a standalone backed table is accepted.

Open destination