Try to pin down what Microsoft Copilot does with your company’s data and you soon meet two phrases that look like rival products. Older Microsoft pages and community threads describe commercial data protection. Newer ones describe enterprise data protection. Neither page mentions the other, and nothing tells you whether these are separate tiers, whether one costs more, or whether your business somehow ended up on the weaker one.
If you have hit that wall, the confusion is Microsoft’s doing rather than yours, and the answer is simpler than the search results suggest. This piece untangles the two names, sets out what the protection covers in plain English, and finishes with the two checks worth running in your own business this week.
Are commercial and enterprise data protection different Copilot tiers?
They are two names for the same promise. Microsoft introduced Copilot with commercial data protection in November 2023 to mark out protected, work-account use of the assistant. In September 2024 it announced enterprise data protection for everyone signed in with a work account, and the rename followed in December 2024. The new label replaced the old one rather than adding a tier above it.
Both terms still circulate because the older material never left the internet. Admin guides, blog posts and Microsoft’s own community answers written through 2023 and 2024 use the earlier name, and many still rank well in search. Jisc’s National Centre for AI, writing for UK institutions, described the September 2024 change as removing the difference in protection levels between free Copilot and the paid Microsoft 365 Copilot, provided the sign-in is an organisational account. The independent tracker M365 Maps records the rename itself. So when you meet either phrase in documentation, you are reading about the same underlying commitment at different points in its history.
What does enterprise data protection actually promise?
Enterprise data protection means the prompts your staff type and the responses Copilot generates are treated as customer data under Microsoft’s Data Protection Addendum, with Microsoft acting as a data processor. They stay within your Microsoft 365 tenant, they carry the same contractual protections as your Exchange email and SharePoint files, and they are not used to train Microsoft’s foundation models.
In practice that means a team member can ask Copilot about a live client situation or a draft proposal without that conversation feeding the model that everyone else in the world uses. The protection also rides on the controls you already have. Copilot respects your permissions, inherits your sensitivity labels and follows your retention policies, so it can only surface content the signed-in user could already open. When it sends a web query to Bing to ground an answer, the query is cut down to a few words and stripped of user and tenant identifiers before it leaves, and those queries are neither shared with advertisers nor used for model training.
The promise has edges worth respecting. Enterprise data protection amplifies whatever state your tenant is in, so wide-open SharePoint permissions stay wide open, now with a faster search tool sat on top. And the ICO still treats your business as the controller under UK GDPR, which means how staff actually use the tool remains your responsibility, whatever Microsoft commits to as processor. None of this replaces proper advice on the addendum itself.
Which sign-ins get the protection, and which do not?
The dividing line is the account. Staff signed in with a Microsoft Entra work account on an eligible Microsoft 365 plan get enterprise data protection, including in the free Copilot Chat experience, at no extra cost. Personal Microsoft accounts, Apple or Google sign-ins, and Copilot inside Microsoft 365 apps for home all sit under Microsoft’s consumer terms, where the enterprise commitments do not apply.
Eligible plans include the ones an owner-managed business is likely to hold already, such as Microsoft 365 Business Standard and Business Premium, so the protection usually arrives with licences you have paid for rather than needing a security add-on. The boundary is still easy to cross by accident. Plenty of people run work and personal accounts on the same device, and the consumer Copilot app happily accepts a personal Outlook.com address, an Apple ID or a Google login. A team member who pastes a client email into that session has moved business data under consumer terms without noticing, and outside your visibility.
One more wrinkle. The Copilot brand covers more than one product, and GitHub Copilot has its own data protection agreement, so enterprise data protection on your Microsoft 365 tenant says nothing about what happens to the code your developers put through it.
Where does your business data go, and what can you audit?
Copilot interactions under a work account are stored at rest in the geography tied to your Microsoft 365 tenant, alongside the semantic index Copilot builds to answer questions. They are logged in the same way as your other Microsoft 365 content, which means they appear in audit, retention and eDiscovery tools rather than vanishing when the chat window closes.
That is a real shift from the earliest version of the promise, which leaned on interactions never being stored at all. Jisc noted that once enterprise data protection is enabled, Copilot activity is logged alongside the rest of the organisation’s Microsoft 365 data. For an owner, the storage is largely good news. You, or the IT partner who looks after your tenant, can search Copilot activity through Microsoft Purview, apply retention rules to it, and check whether client matters are being discussed with the assistant in ways your client agreements would not allow. Consumer sessions never reach those logs, which makes the account boundary an audit boundary too.
What should you actually do about it?
Two checks put the account boundary under your control. First, confirm every member of staff reaches Copilot through their work account, and understands why the sign-in matters. Second, block the consumer Copilot app for business use, or at minimum say plainly that business data never goes into it. Neither check costs anything, and both fit inside an afternoon.
The first check is a conversation. Ask each person to open copilot.microsoft.com and show you which account appears in the corner. A work email on your tenant means the protection applies, and a personal address means it does not, in which case separate browser profiles for work and personal life are the cheapest fix. The second check can be a written policy line in a smaller team, or a technical restriction your IT partner applies to company devices, and Microsoft provides admin controls for exactly this.
There is a bigger safety conversation around Copilot, covering permissions hygiene, sensitivity labels and whether you need a data protection impact assessment. I have covered that separately in is Microsoft Copilot safe for business use. For this question, though, the answer holds. Two names, one promise, and a sign-in boundary that decides whether your business gets it.
