Home

/

Books

/

Keep PII Out of Your LLM

Keep PII Out of Your LLM

Practical strategies for detecting, masking, and never sending personal data to a model.

You-Source

•

September 14, 2026

Contents

1
The Leak You Can't See
2
Blast radius
3
What Counts as PII
4
The Five Doors
5
The Accuracy Reckoning
6
The hybrid that does not work
7
Deterministic Detection
8
NPI, in C#
9
The Three-Way Choice
10
Calling the analyzer from C#
11
Measuring Your Own Demo Gap
12
Choosing the operating point
13
The Ladder of Safeguards
14
Pseudonymisation
15
The Round Trip
16
Restoring safely
17
When Masking Breaks the Task
18
Plausibility hazard
19
The Architecture That Holds
20
The reference architecture
21
Don't Send It At All
22
Structure beats prose
23
The Gateway
24
Failure is a policy decision
25
The Sidecar You Can Trust
26
Egress: deny it at the network
27
RAG and Agents
28
De-identify before you embed
29
Dual-model separation
30
The Boring Controls
31
Evidence, and the First Thirty Days
32
Week two: the chokepoint and the fast layer
33
Entity Catalogue and C# Validators
34
MBI, positional rules
35
Tooling at a Glance
36
Azure AI Language PII, in detail
37
Container Trust Checklist
38
Sources
39
Azure AI Language
40
Provider retention

Want the full PDF to read offline?

Free download — no account required

Free Download

Get the full PDF

All pages, including all code examples, diagrams, and the appendix reference card.
No spam. Unsubscribe at any time.
Your email won't be shared.
Download full PDF
keep-pii-out-of-your-llm.pdf
Oops! There's a problem with your request. We're working on fixing it. Please try again later.