PII detection guardrails
Detect and redact personal information before it reaches a model. Patterns cover names, identifiers, contact details and financial data.
Governance
Data Protection and Privacy
Data protection built into the architecture from the start. Tenant isolation in the database, PII detection in the guardrails and consent-aware processing. Not a feature you turn on - a property of how the platform works.
Tenant isolation
Separated in the database, not just the application.
Every table carries tenant_id NOT NULL with Postgres row-level security. Connector calls and audit entries are tenant-scoped. No query can cross tenant boundaries.
Per-request tenant context
Every API call sets the tenant boundary before any query runs. The boundary is set in the database session, not just in application code.
Row-level security
Postgres enforces isolation at the database level. Even a bug in the application cannot return rows from another tenant.
Scoped audit
Every audit entry belongs to one tenant. The hash chain, evidence and approval records are all tenant-scoped.
PII and sensitive data
Detect, block and redact before it reaches a model.
Guardrails scan inputs and outputs for sensitive patterns. The checks run on every generation step, not once per conversation.
Input scanning
Check every user input for sensitive patterns before the model sees it. Flagged inputs are blocked or redacted depending on policy.
Output filtering
Block responses that contain data they should not. If a model leaks tenant data or personal information, the response does not reach the user.
Consent-aware
Processing respects consent state per data subject. An agent cannot act on data where consent has been withdrawn or was never given.
Regulatory alignment
Controls mapped to the frameworks that apply to you.
The platform carries controls for major data protection regulations. Each set of rules maps to guardrails, audit entries and evidence the platform already produces.
DPDP Act
India's data protection rules mapped to platform controls. Data principal rights, consent management and cross-border transfer conditions.
GDPR and UK GDPR
Data subject rights and processing constraints. Lawful basis tracking, right to erasure and data portability mapped to tenant-scoped records.
CCPA and CPRA
California consumer privacy requirements. Right to know, right to delete and opt-out of sale, enforced through the same guardrail and audit chain.
PIPEDA
Canada's personal information protection rules. Accountability, consent and access principles mapped to tenant isolation and audit controls.
Data residency
Run where your data rules say you must.
Spine is cloud-agnostic. Definitions are portable, state lives in Postgres and the whole platform is containerized. No vendor lock-in on infrastructure.
Cloud-agnostic deployment
Run on AWS, GCP, Azure or your own infrastructure. The platform is containerized and the deploy target is a configuration choice.
Region-aware model routing
Route model calls to regions that match your data residency requirements. Per-tenant policy controls where data is processed.
Portable state
State lives in Postgres. AOPs are git-backed YAML files. Evidence exports to standard formats. Nothing ties you to one cloud.
No vendor lock-in
Moving from one cloud to another is a deploy change, not a rewrite. Your agents, your data and your evidence move with you.
Data protection is architecture, not a feature.
We will show tenant isolation in the database, PII guardrails in action and how evidence stays in your chosen region.