Block ChatGPT, or allow it with controls?
The instinct on seeing shadow AI usage for the first time is to block it outright. That's usually a mistake, and not for a reason that has anything to do with being permissive.
Blocking AI tools outright doesn't stop the exposure — it pushes it to a personal phone or laptop you can't see at all. A four-step sequence works better: warn without interfering first, hard-block only the unambiguous (credentials, government IDs), redact the middle ground (emails, phone numbers), and warn on judgement calls like source code.
Why blocking feels like the safe choice
It's the obvious first move: if a tool is a source of exposure, stop people from using it. A blanket block is fast to roll out, it shows up cleanly in a policy document, and it doesn't require building anything.
What actually happens when you block
The tools are free, they run in a browser tab, and they need no admin approval — which means blocking them on a work laptop doesn't remove the incentive to use them.
If the work laptop won't let someone paste a stack trace into ChatGPT, they'll use their phone, or a personal laptop, instead. You haven't prevented the exposure — the same customer email or the same pasted code still leaves the organization. You've only removed your own visibility into it, and with it, any ability to redact or record what went out.
Blocking also treats every use the same. Most AI usage we see visibility into isn't malicious — it's competent people moving fast, pasting a whole error log rather than the relevant line. A policy built for the rare bad actor ends up punishing the common, harmless case instead.
A sequence that works better
Four steps, in order — each one narrows what still needs a human judgement call.
Run in warn mode first
For a couple of weeks, detect and record without interfering. You learn what your traffic actually contains, and you find your false positives on real data rather than on assumptions.
Hard-block only the unambiguous
Private keys, credentials, government identifiers. Nobody argues these belong in a prompt, so the policy generates no resentment.
Redact the middle ground
Emails, phone numbers, card numbers. The person still gets their answer; the identifier never leaves. This is the highest-value action in the set and the most underused.
Warn on judgement calls
Source code is contextual — internal utility code is fine, client code under NDA is not, and a detector can't reliably tell them apart. A warning puts the decision with the person who has the context.
This sequence, and what a first week of visibility typically turns up, is covered in more depth in what the first week of shadow AI telemetry usually shows.
The realistic goal
Not "nobody uses AI tools." Making the safe path the easy one — allow, with a policy engine deciding what gets through unchanged, what gets redacted, and what gets stopped outright.
See how this plays out by industryFind out what your team is
already sending to AI.
The first week of telemetry usually answers that better than any policy document. No credit card, no commitment — you reach the founding team directly.
Want to see it on your own team before talking price? Join the Design Partner Program — free for 30 days