Skip to content
Atal AI
Enterprise Trust & Compliance

Security engineered into every database query

When you connect store orders, private PDFs, and customer communications, security is non-negotiable. Atal AI enforces forced PostgreSQL Row-Level Security, zero AI model training on customer data, automated PII masking, and AES-256 credential encryption.

πŸ›‘οΈ No credit card required β€’ 5-minute setup β€’ Free tier available

PostgreSQL RLS Engine
Forced Policy Active
Session HandshakeSET LOCAL atal.current_tenant_id = 'tenant_0192a...';
Engine EnforcementSELECT * FROM conversations WHERE tenant_id = current_tenant_id();
βœ“ 15 tenant-scoped tables isolated β€’ Zero cross-tenant data leakage
πŸ”’ AES-256-GCM Encrypted OAuth Tokens β€’ Automated PII Sanitization

100%

Forced RLS

Row-Level Security enforced on all 15 tenant database tables

0

Model Training

Zero customer conversations or documents used for AI training

AES-256

Credential Vault

OAuth credentials and API tokens encrypted with AES-256-GCM

PostgreSQL

Tenant Isolation

PostgreSQL Row-Level Security enforced on all tenant-scoped rows

Security Built From the Database Up

How Atal AI protects customer identities, storefront secrets, and confidential documentation.

Database Isolation

Forced PostgreSQL Row-Level Security

Most multi-tenant apps rely on developer discipline to include `WHERE tenant_id = ...` in queries. Atal AI enforces RLS at the PostgreSQL database engine level with `ALTER TABLE ... FORCE ROW LEVEL SECURITY`. Even an accidental unconstrained query cannot view another tenant’s rows.

  • βœ“PostgreSQL 17 native engine policy enforcement
  • βœ“Automated tenant isolation regression test suite
  • βœ“Non-bypassable even under application code errors
Autonomous Execution Trace

POST /v1/agent/action-dispatcher

βœ“ Security assertion: OK

βœ“ Permission boundary: VALIDATED

βœ“ Transaction status: COMMITTED

Credential Protection

Encrypted Credential & Token Vault

Storefront OAuth tokens (Shopify, WooCommerce, Amazon, eBay, Google, Microsoft) are encrypted before reaching persistent disk storage using AES-256-GCM authenticated encryption.

  • βœ“Unique initialization vectors (IV) per encrypted secret
  • βœ“Automated secret redaction in server log streams
  • βœ“Immediate cascade revocation when a connection is deleted
Autonomous Execution Trace

POST /v1/agent/action-dispatcher

βœ“ Security assertion: OK

βœ“ Permission boundary: VALIDATED

βœ“ Transaction status: COMMITTED

Data Lifecycle

Configurable Retention & GDPR Erasure

Maintain compliance with GDPR and CCPA. Set custom conversation retention policies (e.g. 30, 90, or 365 days) and execute complete customer erasure (right-to-be-forgotten) in a single click.

  • βœ“Automated nightly purging of expired transcripts
  • βœ“Complete GDPR export archives with customer identity graphs
  • βœ“One-click permanent customer data shredding
Autonomous Execution Trace

POST /v1/agent/action-dispatcher

βœ“ Security assertion: OK

βœ“ Permission boundary: VALIDATED

βœ“ Transaction status: COMMITTED

The Secure Data Pipeline

How data is encrypted, processed, and purged across its lifecycle.

STEP 1

Ingestion & PII Redaction

Credit cards, SSNs, and passwords are automatically masked before logging or embedding.

STEP 2

Isolated Execution

Queries execute under tenant-bound database roles with strict RLS enforcement.

STEP 3

Zero-Training Inference

Prompts are sent to inference endpoints with zero-retention and zero-training guarantees.

STEP 4

Automated Purging

Retention jobs automatically delete expired transcripts according to your policy window.

Frequently Asked Questions

Clear answers to common questions about security, pricing, and integrations.

Is my company data used to train AI models?

Never. Neither Atal AI nor our inference upstream providers (OpenRouter, Anthropic, Meta) train or fine-tune models on your uploaded documents, customer inquiries, or conversation transcripts.

How does Row-Level Security (RLS) protect our data from other tenants?

Every database query runs under a dedicated database session with `current_tenant_id()` set. PostgreSQL enforces table policies at the engine level, guaranteeing one tenant cannot read or write another tenant’s rows even if an application bug exists.

How is customer data isolated and protected?

Every database query enforces PostgreSQL Row-Level Security at the database engine level, ensuring isolation between tenants enforced by PostgreSQL Row-Level Security. Credentials and tokens are sealed with AES-256-GCM.

Do you offer a Data Processing Addendum (DPA)?

Yes. Our standard DPA is available online at /dpa and includes EU Standard Contractual Clauses (SCCs) for enterprise compliance.

Enterprise customer service you can trust

Experience true database tenant isolation and citation-grounded answers.

Free tier included β€’ No credit card required β€’ Credentials encrypted (AES-256-GCM)