The short answer
Ungoverned Copilot is Copilot switched on. Governed Copilot is Copilot set up.
Both run on the same Microsoft license, in the same Word, Excel, Outlook and Teams. Switched on, Copilot keeps Microsoft's defaults, so it reads everything each person can open and can use every AI provider Microsoft turns on. Once it's set up, someone has chosen those settings for your firm and kept a dated record of each choice. Ask whoever runs your Microsoft 365 the five questions below, and get the answers in writing. If they can't answer two or more, do the permission review before anyone else gets a license.
What Copilot is
If you're new to Copilot, start here.
§What Copilot does
+
Copilot is Microsoft's AI assistant for work. It sits inside the tools your staff already use. It drafts in Word, works through a spreadsheet in Excel, summarizes a long thread in Outlook and catches someone up on a Teams meeting. It also has a chat window of its own, and staff can build small assistants on top of it, called agents.
The name changed this summer. The work version, until recently Microsoft 365 Copilot, is now called Microsoft Copilot. The same Copilot app now serves personal accounts too, so the name won't tell you which rules apply. That depends on the account a person signs in with. Microsoft says a firm's "security, privacy, compliance, and administrative controls continue to apply when you use Copilot with your work or school account." A staff member signed in with a personal account is outside those controls.
Copilot answers from what the person can already open. Microsoft says Copilot presents "only data that each individual can access." It doesn't reach past a person's permissions. That's usually offered as reassurance, but it's also why governance matters, as the next section explains.
What "ungoverned" means
"Ungoverned" isn't an accusation, and it doesn't mean anyone did something wrong or that Microsoft's product has a flaw. It means Copilot is running on the settings it arrived with. Microsoft wrote them for every business in the world, so they weren't made with a firm that holds tax returns, client matters or account numbers in mind.
§The default settings
+
It reads everything each person can open. In many small firms nobody ever set those permissions on purpose, and they built up over the years. A server moved to the cloud with its folders intact, links were shared with "anyone," and someone who needed one file was given the whole folder. Copilot doesn't create those gaps, but it makes them easy to find with a plain question. Our brief Copilot Does Not Leak. It Reveals. walks through where they come from.
It can use AI providers the firm never chose. Copilot's AI now comes from three places. Microsoft runs GPT models itself. OpenAI runs GPT models as a Microsoft subprocessor, meaning a company Microsoft hires to do part of the work. Those have been on by default for eligible commercial customers since July 24, 2026. Anthropic runs Claude models as a Microsoft subprocessor too, and they're on by default for most commercial customers outside the EU, EFTA and UK. Both outside providers are excluded from Microsoft's in-country processing commitments, which are its promises to run the AI inside a given country.
Nothing knows which files are sensitive. Microsoft's labels and data-loss rules exist, but out of the box none of them know which folder holds client returns.
Staff can build their own agents. Agent Builder comes with the Copilot license unless an administrator limits it.
There's no record. Your firm can search and keep Copilot activity, but only once someone sets that up. Until then it has no record of what anyone decided, or when.
Ungoverned and governed compared
Both run on the same license and in the same apps. Every row below is a setting someone either chose or left alone.
| Question | Ungoverned | Governed | What it means |
|---|---|---|---|
| What it can read | Everything each person can open, overshared folders included | What a permission review leaves open | The partner compensation file stops turning up in answers |
| Sensitive client files | Treated like any other file | Labeled, and kept out of Copilot where the work does not need them | Returns, matters and payroll get a rule, not just a folder name |
| Sensitive numbers in prompts | No rule | A rule that stops Copilot answering when a prompt holds a Social Security number (preview) | A slip gets caught before it becomes an answer |
| Which AI models | Microsoft's own, plus OpenAI's and Anthropic's, both on by default for most commercial customers | Chosen for each group of staff, and dated | The firm decides who uses which provider |
| Outside AI apps | Whatever an administrator has approved, whenever they approved it | A written decision on each one, including Claude's Microsoft 365 connector | One administrator consent can open your whole Microsoft 365: decide on purpose |
| Agents | Agent Builder comes with the license unless an administrator limits it | Reviewed agents that read one library and hand the decision back to a person | An agent that cannot wander cannot wander into a client file |
| The record | None by default | Retention turned on, searched once to prove it works, every setting dated | The firm can answer "show me" |
| When Microsoft changes something | Nobody notices | Checked every quarter | This July alone, a new provider was switched on by default |
What changes when Copilot is governed
What changes when Copilot is governed
In practice, governance comes down to five things, and each one has a before and an after.
Ungoverned (switched on)
- Reads everything each person can open
- Both outside AI providers on by default
- No record
Governed (set up)
- Reads only what the permission review leaves open
- For return work, models set on purpose and the reason written down
- Every setting dated
01What it can read
+
Ungoverned: Copilot reads everything each person can open, and in a small firm that's usually more than anyone intended.
Governed: Someone first reviews who can open what, checked against the people who work there this week rather than an old org chart. After that, Copilot reads only what the review leaves open.
This is the biggest single difference, and it has nothing to do with AI. The review would be worth doing even if Copilot didn't exist.
02Labels and rules
+
Ungoverned: A client return and a lunch menu look the same to the system.
Governed: Client folders carry sensitivity labels designed around the work your firm actually does. A label is a tag that marks a file as confidential. Microsoft Purview, where Microsoft 365 keeps its data protection settings, then keeps Copilot away from labeled files where the work doesn't call for them. When a label encrypts a file, Microsoft says Copilot "honors the usage rights granted to the user." That means Copilot can do with that file only what the person is allowed to do. A second rule, still in preview, can stop Copilot from answering when a prompt contains a number such as a Social Security number.
03Which AI models run, and for whom
+
Ungoverned: Copilot can send work to GPT models Microsoft runs, GPT models OpenAI runs and Claude models Anthropic runs. For most U.S. commercial tenants (a tenant is a business's own Microsoft 365), both outside providers are on by default.
Governed: Your firm decides, group by group. Microsoft lets administrators set both providers for specific users or security groups, so the choice doesn't have to be everyone or no one. A tax firm might keep everyone who handles return data on the models Microsoft runs itself, and might allow Claude for a named group that never touches returns.
Where your data is stored and where the AI runs are separate questions with different answers. Microsoft stores the prompt and the response at rest in the default geography of your firm's Microsoft 365. Where that default geography is the United States, Microsoft commits to keeping them there.
Where the model runs is still open. Microsoft says its own models route requests "to the closest data centers in the region, but also can call into other regions where capacity is available during high utilization periods." Microsoft doesn't yet commit to processing inside the United States for U.S. tenants, though it expects to offer it by the end of 2026.
For a firm under IRC §7216, that open question is the reason for Claremont's position. Until Microsoft commits to U.S.-only processing and the firm can show who reaches the data, client tax return information goes into Copilot only with each client's consent. Our brief Copilot or Claude? The Setup Matters More Than the Tool covers this in full.
04Agents
+
Ungoverned: Agent Builder comes with every Copilot license unless an administrator limits it, so staff can build agents on whatever they can open.
Governed: Someone reviews each agent before anyone uses it. Each one reads a single library, cites the document behind every line and says "not found" instead of guessing. It hands the decision back to the professional and doesn't, on its own, send anything it writes. An assistant kept to one library like that is a smaller risk than a general assistant that can read your whole Microsoft 365.
05The record
+
Ungoverned: If a client, an insurer or an examiner asks what the firm did before turning on AI, the honest answer is "we turned it on."
Governed: Every decision above carries a date and the role of the person who owns it. Microsoft Purview can search and retain Copilot interactions, and in a governed setup someone turns that retention on and tests it. When someone asks, your firm can hand over a record rather than a promise.
Why this matters more for regulated firms
For a firm without client data, ungoverned Copilot is mostly an efficiency question. Regulated firms already work under rules that ask for the things governance produces.
§What the rules already ask for
+
The FTC Safeguards Rule. For the firms it covers, tax preparers among them, it asks that each person can reach only what their job needs, and that someone reviews that access periodically.
16 CFR 314.4(c)(1)(ii)
Limit authorized users' access only to customer information that they need to perform their duties and functions, or, in the case of customers, to access their own information;
16 CFR 314.4(c)(1)(ii)It also asks firms to "monitor and log the activity of authorized users," under 16 CFR 314.4(c)(8). Ungoverned Copilot reads everything each person can open, so in a firm where nobody has reviewed who can open what, that works directly against the access rule. And a firm with no record of its settings can't show the monitoring and logging.
IRC §7216. For tax firms, the question is whether return information is disclosed for anything other than preparing the return, and where the people who receive or view it are. A contractor's employee abroad who only views it puts the disclosure outside the United States. The answer depends on facts such as which AI provider handles a prompt, and where. Governance won't answer the legal question for you, but it gives your counsel the facts to answer it with. Our brief Copilot or Claude? The Setup Matters More Than the Tool covers it in full.
Clients and insurers. They increasingly ask how a firm uses AI. "We use Microsoft" doesn't answer that question, so be ready to hand over a dated record instead.
What governance does not mean
A few misunderstandings come up often enough to deal with directly.
§Four common misunderstandings
+
It doesn't mean blocking AI. The point is to give staff a governed tool that beats the personal accounts they'd otherwise reach for.
It doesn't mean Copilot runs inside your firm. Copilot works within Microsoft's service, under your firm's agreement with Microsoft. The models run in Microsoft's cloud or with the outside providers Microsoft uses, and governance decides which ones, for which staff.
It doesn't make a firm compliant on its own. No software can do that. Governance gives you the settings and a dated record of them, and whether they satisfy a given rule is a judgment for your firm and its advisers.
It isn't a one-time project. Microsoft changes Copilot often. In one week this July it switched on OpenAI's own models by default. It also added a setting that lets non-federal customers in GCC, Microsoft's government cloud, use Anthropic models. That's why a governed setup is checked again every quarter.
How to tell which one you have
Ask whoever runs your Microsoft 365 these five questions, and ask for the answers in writing.
- When did someone last review who can open which files, and is there a record of it?
- Which AI providers can Copilot use in our Microsoft 365, and for which people?
- Are our client folders labeled, and does anything keep Copilot away from them?
- Who can build or share Copilot agents here, and has anyone reviewed the ones that exist?
- If a client or an insurer asked tomorrow how we govern AI, what would we hand them?
If two or more answers are "not sure," you have ungoverned Copilot. That's common, and it can be fixed. Getting the order right matters more than going fast, so do the permission review before anyone else gets a license.
Sources
§Sources+
Microsoft
- Microsoft Learn: data, privacy and security for Copilot page updated 2026-08-18
- Microsoft Learn: Enterprise data protection in Microsoft Copilot and Microsoft Copilot Chat page updated 2026-08-18
- Microsoft Support: Copilot app updates for personal, work, and school accounts posted 2026-09-16
- Microsoft Learn: Anthropic models in Microsoft Online Services page updated 2026-09-18
- Microsoft Learn: OpenAI as a subprocessor in Microsoft Online Services page updated 2026-08-18
- Microsoft Learn: Microsoft Purview DLP for Microsoft 365 Copilot and Copilot Chat page updated 2026-09-17
- Microsoft Learn: Agent Builder in Microsoft 365 Copilot page updated 2026-07-23
- Microsoft 365 blog: in-country data processing for Copilot posted 2025-11-04, editor's note 2026-04-03
Regulation
- 16 CFR 314.4, Safeguards Rule elements eCFR current as of 2026-09-17
Claremont Security helps regulated firms get more done with AI on client work, with the setup and the record to show for it, whichever tool the firm uses. To see where your firm stands before anyone calls, try the AI Readiness Check. It's free and needs no account: claremontsecurity.com/readiness
Get a one-page estimate for your firm: what the work would take and what it’s worth, from your own numbers. No charge.
claremontsecurity.com/insights/governed-vs-ungoverned-copilot
