Compliance monitoring · Production

Building Secure RAG Applications with Realm and ApertureDB

The problem

RAG-based AI chatbots lack native role-based access control, potentially exposing sensitive enterprise data — including personal information, company financials, and intellectual property — to unauthorized users, creating business and regulatory risk.

First attempt

Traditional RBAC systems struggle with scale and flexibility at enterprise level, and link-based document sharing creates gaps where chatbots serve restricted content to anyone who holds the link — a major cause of data leakage.

Workflow diagram · grounded in source
1
Crawl enterprise data sources
integration
“Enterprise data resides in SharePoint, OneDrive, AWS S3, and internal databases. Realm's connectors crawl these sources, discovering: Users (employees, contractors, service accounts), Groups (departments, project teams, security groups),…”
2
Build ACL graph in ApertureDB
integration
“Once the ACL graph is constructed, we persist it in ApertureDB as a structured entity-relationship model”
3
User query submitted
trigger
“When the user makes a query, its embedding is compared to the existing embedding corpus to find the most relevant data which is passed as context to the LLM query”
4
Permission-aware vector retrieval
ai_action
“a RAG query would do a vector search and seamlessly navigate through this graph to filter based on permissions and return only the documents that not only match the vector search but also the querying user's permissions”
Reported outcome

ApertureData claims ApertureDB lowers setup time by 6-9 months, speeds up development by 10X, delivers savings upwards of $2M per team of 10, and performs 2-35X better than contemporary databases.

Reported metrics
Setup time reduction6-9 months
Development speed improvement10X
Cost savings per team of 10upwards of $2M
Performance vs contemporary databases2-35X
Reported stack
RealmApertureDBLLMSharepoint, Google Drive, Slack, GitHub
Source
https://mlops.community/blog/is-your-chatbot-secure
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

ApertureData claims ApertureDB lowers setup time by 6-9 months, speeds up development by 10X, delivers savings upwards of $2M per team of 10, and performs 2-35X better than contemporary databases.

What tools did this team use?

Realm, ApertureDB, LLM, Sharepoint, Google Drive, Slack, GitHub.

What results were reported?

Setup time reduction: 6-9 months; Development speed improvement: 10X; Cost savings per team of 10: upwards of $2M; Performance vs contemporary databases: 2-35X (source-reported, not independently verified).

What failed first in this deployment?

Traditional RBAC systems struggle with scale and flexibility at enterprise level, and link-based document sharing creates gaps where chatbots serve restricted content to anyone who holds the link — a major cause of da…

How is this compliance monitoring AI workflow structured?

Crawl enterprise data sources → Build ACL graph in ApertureDB → User query submitted → Permission-aware vector retrieval.

Related compliance monitoring cases