White paper · Architecture
Index in place: an AI architecture that leaves your documents where they are
Why a read-only, permission-aware index inside your own environment is the safer default for enterprise AI, and what it takes to run.
White papers
For the people who have to sign it off.
Written for architects, security reviewers and data leaders. Each paper explains what K-Lake does and why, in enough detail for your team to check it. No registration, no form.
Four papers
Architecture, security, regulated environments and cost
White paper · Security
Permissions that hold: access control for AI, enforced in the database
How K-Lake enforces file permissions in the database, treats every assistant as a named user, classifies content and records every request.
White paper · Regulated environments
AI behind the boundary: deploying K-Lake where data cannot leave
Three deployment configurations, offline licensing and on-premise extraction: how K-Lake runs where documents must stay inside a defined boundary.
Research note · AI cost
What an answer costs: how K-Lake reduces the tokens an assistant spends
Four retrieval approaches on the same files and questions, the tokens each one used, what that means for an AI budget, and the limits of the test.
Then
Check it on your own data
Every claim in these papers can be tested in a proof of concept on your infrastructure, with your documents. Talk to us about one.