Securing Microsoft 365 Copilot
Keep Personal Data Out of Microsoft 365 Copilot Prompts
Microsoft 365 Copilot sits inside the tools people use all day: Outlook, Teams, Word, Excel and the Copilot chat. With a work (Entra ID) account it runs under Microsoft's Data Protection Addendum, and Microsoft says prompts and responses aren't used to train foundation models. The consumer Copilot app, signed in with a personal Microsoft account, doesn't come with those commitments.
Two risks remain even on work accounts. What people type or attach to a Copilot chat still goes to Microsoft for processing, including attachments with customer data. And Copilot reads what is already in Microsoft 365, so over-shared folders become searchable through it.
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
Microsoft Purview labels and permissions decide what Copilot can read inside Microsoft 365. Maskflare covers the other side: what people paste and attach into Copilot chat, and every other AI tool they use next to it, under one policy.
The Maskflare approach
The Maskflare desktop agent routes Microsoft 365 Copilot chat traffic through Maskflare. Personal data and secrets in prompts and attachments are masked before Microsoft processes them, and the real values are restored in the answer. Consumer Copilot sign-ins can be blocked while work accounts are allowed. The same rules cover ChatGPT, Claude, Gemini and the rest, so a policy doesn't depend on which assistant someone opens. Keep Purview labels and access reviews in place for the data Copilot reads inside Microsoft 365.
Benefits
Prompts and attachments
Masked in Copilot chat before Microsoft processes them, with real values restored in the answer.
Work vs personal
Allow Copilot with work accounts; block the consumer app's personal sign-ins on company devices.
Alongside Purview
Purview governs what Copilot reads in Microsoft 365; Maskflare governs what people send to it.
One policy for every assistant
The same rules for Copilot, ChatGPT, Claude, Gemini and the rest.
Coaching
Warn-and-justify prompts in the moment, with your AI policy link.
Built on
Understand the underlying risks
Frequently asked questions
Does Maskflare work with Microsoft 365 Copilot?
Yes. Prompts and attachments in Copilot chat are masked before Microsoft processes them, and real values are restored in the answer.
Do we still need Microsoft Purview?
Yes, for what Copilot can read inside Microsoft 365: sensitivity labels, permissions and over-sharing. Maskflare covers what people send to Copilot and to other AI tools.
Is Microsoft 365 Copilot GDPR compliant?
No product is compliant by itself. With work accounts it runs under Microsoft's DPA and EU Data Boundary commitments; check the flex-routing and model settings in the admin center. See our Copilot GDPR article.
Can we block the consumer Copilot app?
Yes. Maskflare can block personal Microsoft account sign-ins to Copilot on managed devices while allowing work accounts.
Does masking change Copilot's answers?
Copilot sees consistent tokens instead of the values, and Maskflare restores the real values where the answer repeats them.
What about Copilot inside Word and Excel?
Copilot features inside Office apps read the document you have open. Use Purview labels for those documents; Maskflare covers the Copilot chat and attachments.
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 →