Securing ChatGPT
Use ChatGPT at Work Without Leaking Customer Data
ChatGPT is the AI tool most employees already use, often in more than one way: a company ChatGPT Enterprise or Team workspace, a personal Plus account, the free tier without signing in, and the desktop app. Each comes with different terms. Business workspaces aren't used for model training by default; personal accounts depend on the user's own settings. Check OpenAI's current terms for the plan you use.
Whatever the plan, what people paste still reaches OpenAI: customer emails with names and IBANs, contract drafts, support tickets, code with API keys. Uploads are the bigger risk: a spreadsheet export or a PDF statement can carry thousands of records in one click.
What the employee types
Draft a reply to Ada Lovelace about her
refund of CHF 2,400 to card
4532 7781 0021 9047
IBAN CH93 0076 2011 6238 5295 7
What the AI provider receives
Draft a reply to mf_mask_170b about her
refund of mf_mask_9c2e to card
mf_mask_41af
IBAN mf_mask_d7e0
Why the usual approach falls short
ChatGPT Enterprise settles the contract question. It doesn't stop personal data from reaching OpenAI, and it can't stop people using their personal account for work. Maskflare covers both.
The Maskflare approach
On managed devices, the Maskflare desktop agent routes ChatGPT traffic through Maskflare. Prompts are masked with your rules before they reach OpenAI and the real values are restored in the answer. PDF, Word, Excel and PowerPoint uploads are scanned, images are read with OCR, and answers are inspected on the way back. Maskflare tells your company workspace from personal accounts, so you can allow one and block or strictly mask the other, and warn-and-justify prompts coach people when something sensitive is about to go out. Setup takes three steps: roll out the agent with Intune, Jamf or Group Policy, add your ChatGPT workspace ID in the console, and choose mask, warn or block per data category.
Benefits
Prompts
Names, IBANs, national IDs, health data and secrets are masked before OpenAI sees them, and restored in the answer.
Uploads
PDF, Word, Excel and PowerPoint files are scanned; images are read with OCR. Files with blocked data are stopped.
Answers
Responses are inspected, so sensitive data coming back from connectors or memory is visible in your reports.
Accounts
Allow your ChatGPT Enterprise or Team workspace; block personal accounts or mask them more strictly.
Coaching
Warn-and-justify prompts explain what was found and link your AI policy, instead of a silent block.
Built on
Understand the underlying risks
Frequently asked questions
Is ChatGPT safe to use with customer data?
Only with controls. Even under an Enterprise agreement the data reaches OpenAI. Masking keeps personal data out of the prompt, and account controls keep work off personal accounts.
Does OpenAI train on ChatGPT Enterprise data?
OpenAI states that business workspaces (Enterprise, Team, Edu) aren't used for training by default. Personal accounts depend on the user's data controls. Check OpenAI's current terms.
Can we block personal ChatGPT accounts but allow our company workspace?
Yes. Maskflare recognizes your workspace and treats personal accounts separately: allow, block, or mask more strictly.
Are file uploads to ChatGPT checked?
Yes. Text files are masked in place; PDF, Word, Excel and PowerPoint uploads are scanned and blocked when they contain data your rules block; images are read with OCR.
Will the answers still make sense?
Yes. Masked values are replaced with consistent tokens, and Maskflare restores the real values where the answer repeats them.
Does this cover the ChatGPT desktop app and Codex?
Yes. The desktop agent covers the native app, and Codex is covered by hooks the agent installs and by the gateway.
See it on your own data
Try Maskflare free.
Bring your hardest prompts.
28-day free trial or 10,000 requests, whichever comes first. Point one app at the gateway or enroll a few laptops,
and watch what gets masked. Or talk to us first.
- SOC 2 Type II: In progress
- ISO 27001: In progress
- EU hosting by default · on-premises available
- Prompts and files never stored
- Never used to train AI models
- GDPR · nDSG · DPA available
- Trust center →