24/05/2026
Here's a question worth asking your team this week:
Do they know what data they're allowed to put into ChatGPT or other AI search engines?
Not the policy your business wrote 18 months ago that's filed in a SharePoint folder nobody can find. The actual, working answer they'd give if you asked them at 4pm on a Thursday with a deadline looming.
In our experience, most people don't know. So they make it up. And the line between "helpful AI shortcut" and "that conversation with the ICO you really didn't want to have" gets thinner every week.
The simplest fix we give clients is a traffic light:
๐ด Red: anything you wouldn't email to an external stranger. Client personal data, payroll, draft accounts, board papers, anything under NDA. This stays out of AI tools unless you have a properly contracted enterprise setup.
๐ก Yellow: internal documents, draft strategy, operational data. Approved AI tools only. Never consumer ChatGPT or random browser plugins.
๐ข Green: anything already public. Marketing copy, published reports, generic research. Fair game.
Three colours. Ten minutes to brief. Saves you a regulator's letter, a client's trust, and quite possibly your cyber insurance renewal.
We wrote about this in more detail in this edition of BitBytes, along with the four-part prompt formula that turns your AI tools from a cost line into a productivity one.
Link in the comments.