The training-toggle myth, and what actually still happens to a client's data when you use AI
What accountants get wrong about AI and client data
The eight things compliance actually requires, and the one most firms think is enough. Fifteen minutes to read.
- The one sentence that explains what 'I turned off training' actually does, and doesn't
- A simple three-route framework for deciding what's safe to put in an AI prompt
- The ten-step checklist most firms can start working through this week

Why this guide
Turning off training is not a compliance policy.
Ask an accountant how they keep client data safe with AI and you'll almost always hear the same line: I've turned off training. It's a reasonable thing to believe. It's also, on its own, close to worthless as a compliance position. Turning off training stops one thing: the provider reusing your conversations to train future models.
Everything else still happens. Your prompt still leaves your computer. It's still processed on a server, usually abroad, and retained for a while. If you process client data through AI, you're the data controller and the AI provider is your processor, a defined legal relationship with eight separate obligations attached.
Not fear. Not theory. Just the eight things the law actually asks of you.
What's inside
Five short sections, one clear position on client data.
The eight things data protection law actually requires, in plain English
A three-route framework for deciding what's safe: no client data, anonymised, or identifiable
Why vendor terms change under you, and what to diarise so it doesn't catch your firm out
The ten-step compliance checklist to work through, starting this week
Who this is for
This guide is for you if…
- You've used ChatGPT or Claude on client work and aren't sure exactly where the line is
- You've turned off "improve the model" and assumed that settles it
- You're the partner who'll be asked "are we allowed to do this" and want a straight answer
- You want a framework you can apply to every new AI tool, not just today's list
What accountants are saying
Real changes, not just good intentions.
Worth the month's membership, this one course alone.
Graeme Morris
I thought our team plan was SOC2 rated. It wasn't. This pushed me to build a proper AI usage policy and update our engagement letter.
Rachelle Swanson
I signed up for ChatGPT Business, completed the data processing agreement, and added the ICAEW clauses to our engagement letters.
Bruce Evans
I created an AI usage policy and issued a newsletter to my clients.
Alison Henderson

About Mark
I am a chartered accountant. I have been in this profession since 1988, ran my own firm for a decade, and have spent the last twenty-five years teaching accountants and bookkeepers around the world. For the last three, a great deal of that has been about AI. Privacy and security is, by a distance, the question I am asked most.
Mark Wickersham FCA · Chartered Accountant · Author of the Amazon #1 bestseller Effective Pricing for Accountants · Trainer to 6,000+ firm owners.
Know exactly where your firm stands. Fifteen minutes.
Free guide, no catch. If you want the full 48-page course afterwards, that's inside the AI Community's Premium tier.
