Private AI
All AI models run inside the platform. Customer traffic is never sent to a third-party AI service.
Security and trust
Altoriz sits in front of your applications, so we hold ourselves to the standard we help you meet: private AI, signed policy, full audit trails and clear service commitments.
Protecting customer data
All AI models run inside the platform. Customer traffic is never sent to a third-party AI service.
TLS in transit. Saved destination credentials are encrypted at rest with AES-GCM.
Customers are isolated from each other on the shared edge and in the control plane.
The edge only enforces policy signed by the control plane.
Mask or hash client IP addresses in forwarded logs.
Run the data plane in your own environment so payloads never leave your network.
Safe by design
The data plane never waits for AI, never silently passes traffic, and never makes a change you can't see or undo.
Fail closed. If the edge cannot load a valid policy, it keeps the last good, signed ruleset rather than passing traffic unprotected.
AI out of band. Inline decisions use compiled rules and fast checks. The agent reasons separately and pushes signed policy updates.
Explained and reversible. Every automatic action is logged with its reason and can be undone with one click.
Account security
Service commitments
| Commitment | Target |
|---|---|
| Monthly uptime | 99.9%, backed by service credits of 10%, 25% or 50% below 99.9%, 99.0% or 95.0% |
| Recovery point objective (RPO) | 1 hour or less for a full region failure |
| Recovery time objective (RTO) | 2 hours or less |
| Support response: SEV1 (outage) | 30 minutes, 24/7 |
| Support response: SEV2 (major degradation) | 2 hours, business hours |
| Support response: SEV3 and SEV4 | 1 and 2 business days |
Altoriz is designed to support SOC 2 controls: tenant isolation, an audit trail, signed rulesets and access controls. It is not yet certified, and certification is on our roadmap. We can share our security controls documentation for your review.
Request documentationWe welcome reports from security researchers and treat them as a priority. Read our disclosure policy and send us the details through a secure form.
Report a vulnerability