A plain-language guide // 2026-03-13

AI for business, without the fear

A plain-language primer on how AI handles data, what privacy tools actually do, and how to tell the difference between real security and a sales pitch.


First, a quick word about fear

When someone shows up with a new AI tool, it's completely reasonable to ask "what happens to our data?" That's not paranoia — it's good business sense. The problem is that fear creates an opening for vendors to sell complexity you may not need. Before you evaluate any AI privacy solution, it helps to understand the basics of how AI actually processes information.


Chapter 1: What is a vector embedding?

When you type text into an AI model, the model doesn't read your words the way you do. It converts them into numbers — specifically, into a list of numbers called a vector embedding. Think of it like GPS coordinates, but for meaning.

The word "cat" might become something like [0.82, -0.14, 0.67, 0.03...] — hundreds of numbers that together represent its meaning. "Kitten" would be very similar numbers. "Skyscraper" would be very different ones.

This is how AI finds meaning and context — by measuring the distance between these number clusters. Words that mean similar things cluster together in this "number space." The model reasons by navigating that space.

"Our client John Smith..."
Converted to numbers
Model reasons about it
Answer comes back

The important thing here: the model is working with your original text as plaintext before converting it. That's where the privacy concern lives.


Chapter 2: The actual privacy problem

When your firm sends sensitive data to an AI model hosted by someone else — whether that's Microsoft, OpenAI, or a cloud vendor — that data passes through shared infrastructure. In a regulated industry, the question is: who can see it, and when?

The worry
Sensitive data sits as readable text on a server you don't control
The realistic risk
Logging, model training, data breaches, or subpoenas on the vendor
The regulatory question
Does transmitting this data violate HIPAA, GDPR, or financial privacy rules?

These are real concerns. But here's the thing — they need to be weighed against what's already happening in your organization right now.

If your team is sharing financial data over unencrypted Teams messages, emailing Excel files, or discussing client details on unrecorded Zoom calls — those are larger, more immediate risks than anything the AI model introduces.


Chapter 3: The simple solution nobody talks about

Before reaching for advanced privacy tooling, consider what most compliance-conscious organizations actually do: replace sensitive identifiers with placeholders before the data ever leaves your system.

John Smith, SSN 123-45-6789
Your system replaces
CLIENT_A, ID_001
Sent to AI

The model still understands the context and can reason about the financial situation, the accounting issue, the pattern — it just never sees the actual sensitive values. You keep a local mapping. The AI never touches real PII.

For most accounting and financial workflows, this approach is practical, auditable, and defensible — without adding complexity or degrading model performance.


Chapter 4: What is Stained Glass Transform?

This is where things get interesting — and where the sales pitches start arriving. Stained Glass Transform (SGT), developed by a company called Protopia AI, takes a different approach. Instead of replacing sensitive data with placeholders, it converts your text into obfuscated vector embeddings on your device before sending anything to the model.

Plaintext data
Small neural network on your end
Scrambled embeddings
Sent to AI model

The idea is clever: the embeddings preserve enough semantic meaning that the model can still reason about your data, but the transformation is mathematically irreversible — nobody can decode them back into readable text.

The tradeoffs they don't put in the brochure

What it gives you

  • Plaintext never hits shared infrastructure
  • One-way transformation by design
  • Works with existing model architectures

What it costs you

  • Model can't cite specific details it never saw
  • Obfuscation introduces noise — hallucination risk goes up
  • Significant added complexity and cost
  • No clear regulatory mandate requiring it

You're trading model accuracy and traceability for a privacy guarantee that regulators haven't actually required yet. That's a significant business decision — not a simple checkbox.


Chapter 5: The regulatory reality

Here's what the regulations actually say — and what they don't.

The bottom line: no major regulation currently requires Stained Glass Transform or anything like it. The sales pitch is getting ahead of the law.


Chapter 6: Fix the house before buying the alarm system

Before spending money on advanced AI privacy tooling, ask whether you've done the basics:

Communication
Are Teams, email, and Zoom channels properly encrypted and logged?
File access
Does SharePoint have proper access controls? Who can see what?
Data governance
Do you have a policy for what data can and can't go into AI tools?
Redaction workflow
Is there a simple, consistent process for removing PII before AI queries?

If any of these are "no" — and in most firms they are — adding Stained Glass Transform on top is like installing a bank vault door on a tent.


Chapter 7: Who actually needs this?

To be clear: Stained Glass Transform is not a bad technology. It's a sophisticated solution to a real problem. The question is whether that problem is yours.

The key concept here is threat modeling — asking honestly: who is your adversary, how motivated are they, and what are the consequences of exposure? The answer changes everything.

Accounting firm
  • Accidental exposure is the main risk
  • Adversaries are opportunistic, not targeted
  • Existing channels already leak more than AI would
  • Simple redaction solves the actual problem
  • Regulators don't require it
High-assurance use cases
  • Nation-state adversaries actively trying to reverse-engineer data
  • Shared infrastructure that may itself be compromised
  • Single exposure could be catastrophic
  • Resources to implement and maintain it properly
  • Intelligence, defense, critical infrastructure

A counter-terrorism analyst querying classified signals intelligence on shared government infrastructure? That's the threat model Stained Glass was built for. A CPA firm analyzing client tax returns? It's not even close to the same problem.

Selling a blast-proof door to someone whose windows don't have locks isn't a security upgrade — it's a mismatch of solution to problem. Good technology applied to the wrong threat model is still the wrong call.


So what should your firm actually do?

  1. Audit your existing data flows. Where is sensitive information already exposed today, before AI enters the picture?
  2. Establish a clear redaction or placeholder policy for AI interactions. Simple, repeatable, auditable.
  3. Choose AI vendors who are transparent about their data handling — Microsoft Copilot with proper enterprise licensing, for example, has clear data residency and no training-on-your-data policies.
  4. Document your approach. Regulators want to see that you thought about it, not that you bought the most expensive tool.
  5. Revisit annually. Regulations are catching up to AI fast. What's optional today may be required in two years.

The goal isn't zero risk — it's proportionate, defensible, and actually implemented. A simple policy you follow beats a complex system you don't understand.


DFTZ — Don't Feed The Zombies. kill -9. Clear the stack.
← back