You will be able to check how an assistant handles your data and whether you are allowed to use it for work.
One of Grace's recruiters had been pasting candidates' CVs into a free personal account to draft shortlist summaries. It saved her an hour a day, and she was proud of it. When Grace asked where the CVs went after that, neither of them knew. They did not know whether the chats were stored, for how long, who could see them, or whether they might be used to train future models. Those CVs held names, phone numbers, addresses and salary histories.
Before quality, features or price, there is a more basic question about any assistant you use for work: what happens to what you put into it, and are you allowed to put it there? This lesson shows you where to find both answers.
Every assistant has settings and policies that govern your chats. The details differ between products and change over time, but the questions to ask are the same. Are my chats used to improve the company's models? Can I turn that off? How long are chats kept, and does deleting one remove it? Who at the company can review them, and when?
Some consumer plans, including free ones, may use your chats to improve the company's models unless you change a setting. The setting is usually in a section called data controls, privacy or similar, and the product's privacy policy and help pages explain what it covers. Read them for the assistant you use most. They are shorter than you might expect, and they answer most of the questions above.
Turning off training use is worth doing for most people, whatever they use the assistant for. It is not the same as making a chat private. Chats may still be stored and reviewed for safety or legal reasons, as the policy will explain, and the temporary chats from lesson 6.3, Memory: what it keeps and how to manage it, have their own rules.
Business and enterprise plans usually come with stronger data terms. Typically, the provider agrees not to train its models on your organisation's data, and the organisation gets controls over retention, access and who can use which features. The terms are set out in a contract between the provider and your employer, not just in a setting an individual can change.
That is why many companies approve only the business version of an assistant, even when the same product is available free. From the company's point of view, the free version may be fine for an individual's personal questions, but it gives the company no control over its data and no agreement it can rely on. Organisations in Singapore also have obligations under the Personal Data Protection Act, the PDPA, for how they handle personal data such as a candidate's CV, which makes them careful about where it goes.
Grace checked. Her agency had a business plan for one assistant, approved by management the year before, and almost nobody in her team knew about it.
For work data, your employer's AI policy is what decides which tools you may use, whatever any tool's own settings say. A personal account with training switched off is still a personal account. If your company's policy says client data may only go into the approved business assistant, then that is the rule, even if another tool's settings look just as safe to you.
Find the policy before you need it. It may be called an AI policy, an acceptable use policy for generative AI or something similar, and it may sit with IT, legal, compliance or HR. If you cannot find one, ask your manager or IT team directly. "No policy" is rarely the true answer. More often there is one that nobody has circulated.
When you read it, look for three things: which tools are approved, what kinds of data may and may not go into them, and whether anything needs approval first. Many policies also say what to do if you have already pasted something you should not have.
Grace's agency policy turned out to be one page long. It said candidate personal data could go only into the approved business assistant, that client contracts could not be uploaded to any assistant at all, and that personal accounts were off limits for anything work-related. Her recruiter moved her CV summaries across that week, and the work still saved her an hour a day.
The finer question of what not to paste into an assistant, even an approved one, is covered in depth in AI at work: writing, research, spreadsheets and meetings. The short version from this course is the test from lesson 1.2, Give context the way you would brief a colleague: if a detail would not change the output, it does not need to be in the prompt.
Knowing your own rules takes about twenty minutes, and you only need to find them once. Open the settings of the assistant you use most, and have your company intranet or staff handbook open in another tab.
Find and read the data settings of the assistant you use most and your company's AI policy, and write down the three rules that apply to you.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).