Zero trust.
Tenancy isolated.
Your intelligence layer must remain private. At Atlas, we architect our databases, pipelines, and model APIs with standard multitenant isolation principles, keeping your research safe and under your control.
Data Partitioning & Multitenancy
Your workspace database records and vector database indexes are logically partitioned. An account can never read or query the embeddings of another tenant. Scope validation checks run on every single query handler.
Zero Model Training Policy
We do not train language models on your files. The text extracted from your PDF, docx, CSV, or md documents is sent to inference endpoints strictly as temporary context, and is never saved by our foundation model partners.
Encryption at Rest & In Transit
All network connections to the Atlas client and API require TLS 1.3 encryption. Files stored in our knowledge buckets are encrypted using AES-256 with rotation schedules managed by secure key storage.
SOC 2 Type II Audited
Atlas is built on top of cloud infrastructures audited for compliance with SOC 2 Type II, ISO 27001, and HIPAA frameworks. We adhere to rigorous physical and logical security compliance parameters.
Data Handling Framework
Where is my document data stored?
All document chunks and database records are hosted in secure, regional clouds located in the United States and India, ensuring low latency and compliance with geographical data sovereignty regulations.
Who has access to my index queries?
Only verified users who possess active tokens for your specific workspace can query or retrieve document references. Atlas support personnel cannot view or search your records unless you explicitly authorize access for a troubleshooting ticket.
How can I request document deletions?
When you delete a file in the document library tab, Atlas immediately schedules it for deletion. All original texts, database indexes, and vector embeddings corresponding to that file are wiped completely from our disks within 24 hours.