01Security
Control is part of the architecture.
Lattice exists because an AI system that can act needs the same controls as any other actor in the enterprise: an identity, a permission boundary, an approval path and an audit trail.
No certification is claimed on this page.
16Security
Control is part of the architecture.
Identity decides who the actor is. Policy decides whether the action may happen. The audit record makes both reconstructable. None of that is a setting applied afterwards.
01
Identity
02
Policy
‖ evaluate
03
Agent
04
Tool
‖ permission
05
Action
06
Audit
‖ record
- 01EncryptionTransport encryption in flight and encryption at rest for stored telemetry and records.Designed
- 02Tenant isolationWorkspace-scoped data paths, with logical isolation as the default and private deployment as an option.Designed
- 03RBACRoles for engineers, approvers, auditors and administrators, scoped per environment.Designed
- 04SSOEnterprise identity through SAML or OIDC, with group-to-role mapping.Designed
- 05Audit logsAppend-only records for agent actions, policy changes and approval decisions.Designed
- 06Secrets managementProvider credentials and API keys held in a secrets manager, never in telemetry.Designed
- 07Data redactionField and pattern redaction applied before payloads are stored or sent to a model.Designed
- 08Retention controlsPer-workspace retention windows for traces, payloads and audit records.Designed
No certification is claimed on this site. Security documentation is available on request.
03Identity
Every action has a named actor.
- Agent identity
- An agent is a registered principal with an owner and an environment, not an API key in a script.
- Human identity
- Console users authenticate through SSO; approvers are named in the audit record.
- Machine identity
- Scoped API keys per environment, with overlapping rotation windows.
- Roles
- Engineer, approver, auditor and administrator, scoped per environment.
- Impersonation
- Not supported. A decision is always attributed to the identity that made it.
04Data protection
The payload is the risk.
- Redaction at ingest
- Field and pattern rules run before anything is written or forwarded to a model.
- Encryption
- Transport encryption in flight; encryption at rest for telemetry and records.
- Separation
- Audit records are stored apart from span payloads and can outlive them.
- Retention
- Per-workspace windows for traces, payloads and audit records.
- Training
- Workspace telemetry is not used to train models.
05Policy enforcement
Enforcement happens before the action.
- Placement
- At the boundary the agent crosses, not inside the prompt.
- Actions
- allow · deny · require_human_approval · redact.
- Precedence
- Deny beats approval, approval beats allow. The matched rule is reported.
- Unavailability
- Fail-closed or fail-open per policy; either way the event is recorded.
- Change control
- A policy change is a new version, and the version is named in every decision.
06Tenant isolation
Workspace is the boundary.
- Logical isolation
- Workspace-scoped data paths on every query, as the default posture.
- Environment isolation
- Production keys cannot read staging telemetry, and the reverse.
- Regional residency
- Data residency selected per workspace.
- Private deployment
- The control plane in your own cloud account — on the roadmap.
- Access by PSIFI
- Break-glass access is designed to be explicit, time-boxed and recorded.
07Enterprise controls
What an enterprise will ask for.
- SSO
- SAML or OIDC with group-to-role mapping.
- SCIM
- Directory-driven provisioning — roadmap.
- Audit export
- CSV and JSON export of the ledger, and streaming to your SIEM.
- IP allow lists
- Restrict console and API access by network — roadmap.
- Data processing terms
- Available on request.
Audit
Reconstruct any action afterwards.
Five fields answer the question an auditor actually asks: who, what, when, why and result. Expand a record to see them aligned, with the policy version and the approver that applied.
who
agt_customer_resolution
what
payments.refund.create
when
2026-01-14 09:41:10.402
why
P-014 v4 · approved by finance_approver
result
rfd_0d93aa
09Security roadmap
What is not true yet.
A security page is only useful if it is candid about its gaps.
- Independent assessment
- No third-party audit has been completed. When one is, it will be named here with its date and scope.
- Certifications
- None held. This site shows no SOC 2, ISO or similar badge, because there is nothing to show.
- Penetration testing
- Planned before general availability; the summary will be available under NDA.
- Private deployment
- Designed, not built. Listed as roadmap on the platform page.
Report a security concern: security@psifi.systems
Put a control layer around your AI systems.
Connect your agents, understand their behaviour, define their boundaries and operate AI systems with a clear audit trail.