Safe by construction, not by policy.
Critical facilities cannot depend on an AI model behaving well. Aidni's safety comes from what agents are unable to do, enforced by the interface itself.
Six layers between an agent and your facility.
Read-only by default
Connectors to EPMS, BMS, DCIM and schedulers read data. They do not write to site systems.
Two-verb agent surface
Agents can Observe and Propose. There is no tool for direct actuation, so no prompt can make an agent act alone.
Human approval
Every proposal names its risk, rollback and approver, and moves through your existing change process.
Execution stays with your systems
Approved changes are carried out by the scheduler, BMS or work-order system that already owns them.
Hard exclusions
Fire suppression, EPO and safety interlocks are permanently outside the agent surface, regardless of configuration.
Full audit trail
Every agent read, proposal, approval and outcome is logged and exportable.
Your facility data stays yours.
| Area | Commitment |
|---|---|
| Tenancy | Single-tenant environments only. No customer data is pooled. |
| Model training | Customer data is never used to train shared models. |
| Residency | Aidni cloud, your VPC, or on-prem and air-gapped under Portfolio. |
| Encryption | Encrypted in transit and at rest. |
| Access | SSO, role-based access and scoped agent policies per hall, row and system. |
| Compliance | SOC 2 Type II is on our roadmap. We will publish status here, not before it is real. |
Security questionnaires and architecture reviews are available to customers under NDA.
Bring us one site. We will show you its real capacity.
A six-week Site Assessment maps the power envelope, the cooling design and the compute plan for one facility, delivered by the engineers who build the platform.