$MAIGA · geopolcoins.xyz
AI

How to stop employees pasting company data into ChatGPT

10 November 2026 · 1 min read · Tool: Human Override

Also in: Nederlands Français

Quick answer

Write a short AI policy, choose which AI tools are approved, and block the rest on work devices and the office network with browser policies and DNS or firewall rules. Human Override's free Pro pack has ready-made blocklists for these.

An employee pastes a client contract into a chatbot to “summarise” it. Nobody meant harm, but confidential data has now left the company. For courts, police and government services, case files and witness statements must never end up in a public AI service at all.

1. Decide what is allowed

Make a list of approved AI tools (for example an enterprise version with a data agreement) and say clearly what may never be pasted anywhere: client data, contracts, source code, case files, personal data.

2. Block the rest on work devices

3. Block it on the network

Load the AI domain list into your DNS server (BIND RPZ, Unbound, dnsmasq, Windows DNS policies), your firewall or your proxy. Every device on the office network then gets “does not exist” for public AI services.

4. Check and keep it current

New AI services appear every month. The Human Override list is public and updated; download the latest Pro pack when you review your rules.

Blocking is one layer, not the whole answer. Data protection rules (such as the GDPR) still apply, and people need training and a clear policy. This guide is not legal advice.

Get the free Pro pack

Questions

Is the Pro pack really free?

Yes. Organisations can ask for paid help with roll-out or a custom list.

Will Microsoft 365 or Google Workspace stop working?

No. Only AI-specific addresses are listed; protected everyday domains such as microsoft.com and google.com are never blocked.

Does it work for courts and government?

Yes. The files are standard formats that government IT can review and roll out.

Read next