What kinds of work details are safe to use in AI tools?

Published: November 28, 2025

Dear 404: What kinds of work details are safe to use in AI tools?
A curious character in the shape of a brain

Dear 404,

What kinds of work details are safe to use in AI tools, and where do I need to draw the line?

— Concerned About Clicking “Submit”

Dear Concerned About Clicking “Submit,”

Pull up a chair and let me give you the real talk about what you should and should not toss into those ever-hungry AI tools.

Think of AI like that helpful coworker from 1997 who’s great at making templates but absolutely cannot be trusted with office gossip, HR files or the location of the spare office keys.

Use it wisely and it’s your best friend. Feed it the wrong thing and suddenly you’re starring in a workplace PSA about data privacy breaches (and no one wants that).

Here’s the scoop!

What you can safely share with AI (aka: “Totally fine”)

Step one: ask yourself what kind of AI tool you’re using. Is it a university-provided tool (like Copilot) or is it something you stumbled across in the wilds of the internet? University tools are security-checked and privacy-reviewed. Random AI tools… not so much. And Dear 404’s golden rule? Never feed university data to an unvetted tool. Ever.

  • Generic drafting & editing tasks: If you’re using the enterprise Copilot (with the green shield displayed), you’re safe to share internal (Level 2 and Level 3) information — your everyday work content. Emails, summaries, messy first drafts… go for it! Just remember even in enterprise Copilot, keep personal information out of it.

    AI is great for polishing your prose, untangling your thoughts, and helping you brainstorm but the world is still learning about how it affects privacy, so personal information is a no go.

  • Anything already public or meant to be shared publicly: Blog posts, public reports, conference intros, information already out there in the world — this is all fair game.

  • Low-stakes administrative or operational odds and ends: Drafting generic communications, formatting documents, taking meeting notes without private details, summarizing high-level discussions… all good. As long as it doesn’t include personal or sensitive info, you can let AI handle the boring bits.
  • Learning, planning, brainstorming and general idea-shaping: Thinking out loud? Planning a project? Building an outline? AI is like that one friend who always has 17 suggestions and a colour-coded list ready. Great for ideas, safe for ideation and perfect for the “I need to organize my brain” moments.

Where to draw the line (aka: “Nope — not here”)

  • Anything with personal information, (SINs, performance reviews, HR issues, student info) or anything that makes the privacy office flinch — leave it out.
  • Level 4 data or confidential university data — the kind only a small group should ever access; internal strategies, unpublished research data.

Dear 404’s (other) golden rule:

Here’s the deal: if something feels even a little “should I really be typing this in?” — don’t. AI tools are for the harmless, everyday details, not the private or confidential ones. Just remember, it’s a helper, not a replacement — we’re talking augmentation, not automation.

Check out U of T’s Guideline on the Use of Artificial Intelligence by University Administrative Staff if you need or want a deeper dive.

If you want to learn more about the GenAI tools, take a look at this list of vetted and supported tools from the university.

Sincerely,
4[0‿0]4