Security & Privacy

Your Clients' Data, Protected by Design

I build websites, automations, and client portals for businesses that can't afford a data problem — law firms, medical practices, financial advisors, HR teams. Security isn't a feature bolted on at the end; it's the architecture everything is built on. Here's exactly how your information is protected — in plain English, no jargon.

Looking for the legal side — privilege, work product, and the 2026 case law? Read the AI & Law analysis →

The One Thing That Matters Most: Isolation

Most popular software runs every customer through one big shared system. Your data sits in the same database as thousands of strangers, separated only by a software setting. We don't do that. Every client gets their own isolated stack — their own database, their own application, their own AI key. Your information never shares a home with another business's.

What "Your Own Stack" Actually Means

When I build for you, your system is provisioned as a self-contained environment. There is no shared, multi-tenant database where a single bug or misconfiguration could spill one firm's files into another's view. The separation is structural, not just a permission checkbox.

Your database
·
Your application
·
Your AI key
·
Your private code repository

It's the difference between a safe-deposit box that's yours alone and a shared filing cabinet the whole building can open. For a business with confidentiality obligations, that distinction is the entire ballgame.

Two Ways to Run It — Both Secure

Every client is isolated. The choice is whose infrastructure that isolated environment lives on. Both options are built to the same security standard; the difference is how much you want to own directly.

Maximum Control

Self-Hosted

Prefer to own every piece? The entire system can be built on your accounts — your database, your hosting, your AI key — so your data never touches my infrastructure at all. You hold every credential.

  • Every account and key belongs to you, end to end
  • Your data lives only on infrastructure you control
  • No third-party hosting relationship to vet — there isn't one
  • I can build on managed infrastructure, then hand the whole thing over to your accounts
Best for: the strictest compliance, privilege, or regulatory requirements where "we own it outright" is the standard.

The Three Tiers of AI Risk

Not all "AI" carries the same risk to confidential data. The difference between dangerous and defensible isn't the brand of AI — it's the terms the data travels under and the infrastructure it runs on. Here's how the options stack up.

High Risk

Public / Consumer AI

ChatGPT Free & Plus, Claude.ai Free/Pro, Gemini consumer

  • Data may be used to train future models
  • Multi-year retention possible
  • Counts as third-party disclosure

Never with confidential data.

Conditional

Shared Cloud Platforms

Multi-tenant SaaS automation — Zapier, Make.com

  • Data flows through shared vendor servers
  • SOC 2 / DPA available on paid tiers
  • Vendor staff have controlled access

Fine for non-privileged work — admin, intake, scheduling.

Defensible

Isolated Infrastructure + Commercial API

Your own private server, or an isolated cloud stack — paired with the AI API on commercial terms

  • Data stays inside an environment dedicated to you
  • AI calls under commercial terms — no training, short retention
  • You own the logs, the data, and the deletion policy

The architecture for confidential work.

Everything I build for confidential work lives in that green tier — whether it's a dedicated private automation server or an isolated portal with its own database. The common thread is always the same: your data stays in an environment that's yours, and the AI only ever runs under commercial terms that forbid training.

How I Use AI — and Why Your Clients Are Covered

A fair question: if I use an AI assistant like Claude to build your system, does your client data end up in it? No — and the reason is simple once you separate building the system from running it.

Think of a contractor who builds a safe for a bank. They use their own tools — drills, welders, measuring tape — to build it. Those tools never touch the bank's money. The money only goes inside after the safe is built and installed.

🔨
Building the system
An AI assistant helps me write the code — the way a carpenter uses a saw. No client information ever goes into that. It's just construction.
🔒
Running the system
When your system is live and handling real client files, it runs on the secure, isolated setup built for you — using the AI API under commercial terms (no training, short retention, your own key). That's the finished, locked safe. Client data only ever flows through there.
AI assistantBuild phase — no client data
Your isolated system + AI APIRun phase — commercial terms

One question settles it: did any real client information pass through the build tools? The answer is no — and I keep it that way. See the full legal analysis →

What Actually Protects Your Data

The specifics, in plain terms. Every system I build for confidential work includes these by default — not as paid add-ons.

🧱

Per-Client Isolation

Your own database and application. No shared, multi-tenant storage where data can cross between clients.

🔐

Encrypted In Transit & At Rest

All traffic is HTTPS-only — unencrypted access isn't possible. Stored data and files are encrypted on disk.

🪪

Authenticated Access + Database Rules

Only the staff you authorize can log in, and access rules are enforced inside the database itself — not just in the app.

🗝️

Secret Keys Stay Server-Side

Powerful keys live in protected server settings — never shipped to the browser, never stored in code.

🧹

Data Minimization

Sensitive uploads go to private storage, and intake documents are deleted right after the needed details are extracted. Less retained means less exposed.

🚫

No AI Training on Your Data

Your data runs through the AI under commercial terms that contractually prohibit using it to train or improve models.

🏛️

Audited, Reputable Providers

The underlying infrastructure is run by providers that maintain independent SOC 2 security audits — with automated backups and redundancy.

🎯

Smaller Attack Surface

No always-on server sitting open to the internet to be broken into. Components run on demand, then stand down — there's less to attack.

Straight Answers

The questions clients actually ask — answered honestly, including the limits. No overpromising.

Can you see our data?
To build and maintain your system, I have the same access any IT provider or hosting partner has — that's what lets me support you. But your data is never sold, mined, or fed into AI training, and access is logged. If you want a setup where no outside party can access your data at all, that's exactly what Self-Hosted delivers.
Is our information used to train AI?
No. Your data runs through the AI on commercial API terms that contractually forbid using inputs to train or improve the model — a different world from consumer AI apps, where training is often the default.
Where does our data physically live?
In your own isolated database on audited cloud infrastructure (region can be selected) — or, on Self-Hosted, entirely within your own accounts. Either way, it's never commingled with another client's.
What happens if we ever part ways?
What's built is yours. Your data can be exported and the system handed over to your own accounts. You're never locked in or held hostage to keep the lights on.
Built for Confidential Work

Let's Talk About Your Setup

Whether you want a fully managed system or to own every key yourself, I'll walk you through exactly where your data lives and how it's protected — and give you documentation you can hand to a client, an auditor, or a compliance officer.

Free templates for your practice

Data Processing Agreement AI Use Policy Template Engagement Letter Language