Accountable AI, by design.

Your data stays isolated

• Your organisation's records are kept apart from every other organisation's. The data layer scopes queries on organisation-owned records to your organisation, and our engineering rule is that every single-record lookup is checked against your organisation.
• Record data stays in your systems. SmixAI reads it under the permissions you grant and keeps only the configuration data and findings it needs.

Your credentials are protected

• Credentials for your connected systems are encrypted at rest with AES-256-GCM and never stored in plain text.
• Platform secrets live in a dedicated secrets manager and are injected when a service starts — never written into code.
• Every proposed code change is scanned for committed credentials.

Agents stay under your control

• Every new agent starts in shadow mode: it proposes, a person handles the case for real, and SmixAI scores how close the proposal was.
• An agent goes live only when an administrator in your organisation promotes it — and only after it reaches its run count and accuracy target.
• By default an agent may only propose: a write to a connected system waits for a person's approval. A write that is not allowed is blocked, sent to a person and recorded.
• Actions and refusals are written to an append-only, tamper-evident audit log.

AI and your data

• Personal details are masked before text is sent to an AI model.
• Each task is routed to a model suited to it, rather than the most powerful model for everything.
• Early accessChoose which AI providers and models your organisation allows, and SmixAI routes only to those.

Access

• Sign-in and organisations are handled by a dedicated identity provider.
• Role-based permissions: owner, admin, member and viewer.
• Only owners and admins can manage billing.

Sub-processors

The providers that process data for SmixAI — what each one processes, and where — are listed on our sub-processor page.