The honest answer is that it depends less on the tool than on the account. The same chatbot can be a reasonable place for a draft contract or a bad one, depending on whether your employee is signed in with a personal account or one your company manages.
Most offices haven’t sorted that out yet, and the people doing the typing usually don’t know there’s a difference.
What actually happens to what you type
When someone pastes text into an AI tool, it goes to the company that runs the tool, gets processed on their servers, and an answer comes back. What happens after that is set by the terms of the account.
Free and personal accounts on the popular AI tools are generally consumer products. Depending on the tool and the settings, conversations can be kept, reviewed, and used to improve the provider’s models, and it’s up to the individual user to find the setting that turns that off. Business plans are a different contract. Most of them commit not to train on your data by default, and they give an administrator control over who has access. Microsoft’s Copilot Chat, for example, signed in with a work account, comes with Microsoft’s enterprise data protection: prompts and responses are covered by the same terms as your company email and aren’t used to train the underlying models.
So when you talk to your staff about AI, ask which account they sign in with, as well as which tools they use.
Training is the smaller problem
People worry most about their data ending up inside a model. That’s a fair worry on a personal account, but day to day, the bigger problem is control.
If an employee does company work on a personal AI account:
- you can’t see what’s been shared
- the chat history, with whatever was pasted into it, stays in their personal account
- when they leave, it leaves with them, and there’s nothing to switch off
- nobody has set multi-factor sign-in on it, so if their personal password leaks, so does that history
You’ve probably spent real money on firewalls, email filtering and multi-factor sign-in for your company accounts. A personal AI account sits outside all of it.
When you talk to your staff about AI, ask which account they sign in with, as well as which tools they use.
How common it is
BlackFog surveyed 2,000 employees in the US and UK in November 2025, all at companies with more than 500 staff (BlackFog, January 2026). Almost half said they use AI tools their employer hasn’t approved. Of those, 58% use free versions. Among all respondents, 27% said they had put employee data such as names, pay or performance information into unapproved tools, and 23% had put in financial statements or sales data.
Those are bigger companies with IT departments. A 20-person office with no AI rules at all is unlikely to be doing better.
What a safe setup looks like
You don’t need to read every vendor’s privacy policy to get this right. A setup is in good shape when:
- Staff use company accounts, not personal ones. Signed in with their work email, so access ends when their job does.
- The terms say your data isn’t used for training, and ideally that prompts aren’t kept after the answer comes back.
- Sign-in uses multi-factor authentication, the same as your email.
- Someone can see who is using it and remove access in one place.
- You decide what it can connect to. AI tools that can read your files or act in your systems should only reach what you’ve chosen, and sensitive actions should wait for a person to approve them.
- There’s a written policy that says which tools are approved and what never goes in. We cover that in What to Put in an AI Use Policy.
The realistic options for a small business
For most small businesses it comes down to three choices.
Business plans of one AI tool
Pick one, buy business seats, and turn off personal use for work. Simple, but you’re tied to one company’s models.
Copilot inside Microsoft 365
If you already run on Microsoft 365, Copilot Chat with work accounts is the shortest step, and it sits under the agreement you already have with Microsoft. The paid Microsoft 365 Copilot add-on goes further and works across your own email and files.
A managed AI workspace
One place, run for your company, that gives staff access to several models with the controls above built in. This is what we provide.
If you don’t choose any of them, the work carries on in personal accounts, where you can’t see it.
What we offer
Our Secure AI workspace gives your team chat, document work and mobile apps with models from OpenAI, Anthropic and others, in an environment your business controls. It runs on Microsoft Azure, each business in its own space. Model calls are made with zero data retention, and the platform doesn’t train on your data. You decide who has access, which models they can use and what the AI can connect to. If you’d like to see how it handles your data before you decide anything, call us at (386) 252-2292 and we’ll walk you through it.