You will be able to explain how consumer and business AI accounts differ in how they store and use your data.
Before you paste work material into an AI assistant, find out what your own account does with it: whether your words train future models, how long they are kept, who may read them and how to delete them. The answers sit in your account's settings and the provider's terms, and they vary from one plan to the next.
Hafiz, an invented example, does HR and admin at a forty-person engineering services firm in Jurong. Last month a staff complaint landed on his desk. He copied the whole thing into the free assistant on his phone, with the staff member's name and the details of the dispute, and asked for help drafting a calm reply. What came back read better than anything he could have written himself that afternoon. Travelling home on the MRT, he began asking himself where the complaint had ended up.
A chat with an assistant can feel like a private exchange with a program on your own screen, though that impression is wrong and most users never think to ask where their words travel. The assistant does not run on your phone or laptop. The model lives on servers run by the provider, and anything you type or upload is transmitted to those servers for processing, including attached files, images you share and, with some apps, your spoken words.
After that, it comes down to which provider and which plan you're on. As a general pattern, conversations are stored for a while, which can mean until you delete them or for a set period afterwards. Many providers reserve the right for their employees or contractors to read conversations in certain situations, such as looking into abuse or working on safety. How long your data stays and who can see it are set out in that provider's terms.
Once you've sent something, your control over it is weaker than your control over a file saved on your own drive. Removing a chat clears it from your list of conversations, while what happens to the provider's copy, and when, is governed by their terms.
In lesson 8.2 of the course "Working with AI assistants", "Data settings and the rules at your workplace", you met the difference between consumer and business plans. The task now is to apply that difference to the account you actually use.
A single assistant is often sold in several forms: a free plan, a paid plan for individuals, and plans for companies. The model underneath may be identical across them, yet the terms usually differ. Consumer plans are the free and paid personal versions. With some of them, the provider may use what you type to train or improve later models unless you turn that off in your settings. Certain apps include a switch for this, and some offer a temporary or incognito chat, a mode that keeps the conversation out of your history. Defaults vary between providers and shift over time, so your own account is the place to check, whatever a colleague says about theirs.
Business and enterprise plans are sold to companies and usually differ from consumer plans on two points. The contract generally commits the provider to not training models on the company's data, and the company has an administrator who controls settings, retention and access. Assistants inside office software, for example Copilot in Microsoft 365 or Gemini in Google Workspace, are typically covered by the company's own agreement with the provider, provided the company has set them up under a work account.
For this reason, "Is it safe to paste this into AI?" cannot be answered once for every case. A client contract pasted into an assistant your company manages may well be allowed. The same contract pasted into a free personal account on your phone almost certainly breaks the rules.
Providers revise their terms, introduce new settings, rename plans and alter defaults, so a colleague's summary or a blog post written a year ago may describe rules that have since changed. Begin with the source: the privacy or data settings inside the app you really use. Then look up the provider's privacy policy and, if you are on a business plan, the data or trust page for that particular plan, where the provider explains how that plan handles data. Use them to answer three questions:
Are my conversations used to train models, and can I turn that off? How long are they kept, and what happens to them after I delete them? How do I delete them, and does deleting them also remove files I uploaded?
If your company provides the assistant, your IT team or data protection officer will usually answer these faster than the provider's website will. Asking them also tells you whether your company has an AI policy in the first place, which is the subject of lesson 2.3.
Checking the app, Hafiz discovered it was a free personal account. The option allowing his conversations to be used for model improvement was on, because he had never opened that screen. He turned it off and deleted the chat. Then he asked his manager what the firm permitted, and learned that the firm had a business plan for the very same assistant, though nobody had ever set him up on it.
After that, staff matters went into the business account only, and only once they had been redacted, a skill you practise in lesson 2.4. The order he followed works for you too: confirm which account and plan you are on, switch off model-improvement use, delete the chat, ask what your workplace allows, move sensitive work into the company's business account, and redact before you paste.
He had not done serious damage. If anyone had asked Hafiz what had become of that complaint, though, he would have had no answer, and for someone in HR that is an uncomfortable place to be.
Check your own account the same way now. Open the assistant you use for work, find its data and privacy settings, and learn what it does with your text before you type anything else into it.
Find the data and privacy settings of the assistant you use for work and note whether chats are used for training, how long they are kept and how to delete them.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).