Understand How Your Data Is Used
Most free, consumer-facing AI chatbots use your conversations to train their future models. This means anything you type—from a simple question about historical dates to a draft of a personal essay—can become part of the system's knowledge base. In some
cases, this has led to sensitive information being inadvertently exposed. For instance, a bug in ChatGPT once revealed the chat titles of other users. Before you start using any AI tool, assume your conversations are not private by default. The fundamental mechanism of these platforms is data collection, which helps them refine their services. This isn't necessarily malicious, but it creates a risk you need to manage.
Never Share Personally Identifiable Information
This is the most critical rule for protecting yourself. Avoid entering any data that can be used to identify you or others. This includes your name, address, student ID number, financial information, or specific details about your school or medical history. Even seemingly harmless details, when combined, can create a detailed profile. Universities and privacy experts are clear on this point: never enter any information into a public AI tool that is protected by privacy laws like FERPA, which governs student educational records. A single accidental paste of a class roster or a graded paper with a name on it can constitute a data breach with no way to undo it.
Use Privacy Settings and Opt-Out Features
Fortunately, most major AI platforms now offer controls that let you limit how your data is used. For example, both ChatGPT and Google's Gemini have settings that allow you to turn off the feature that uses your conversations for model training. In ChatGPT, this is typically found in 'Settings' under 'Data Controls,' where you can toggle off the option to 'Improve the model for everyone'. In Gemini, you need to visit your Google 'My Activity' page and disable 'Web & App Activity', as well as turn off 'Gemini Apps Activity' in Gemini’s own settings. Taking a few minutes to configure these settings is one of the most effective steps you can take.
Generalise Your Prompts
You can get help from an AI without giving away sensitive context. Instead of pasting in a full, confidential assignment prompt from your professor, generalise the query. For example, instead of asking the AI to 'Critique my essay on the economic impact of the 2024 Mumbai port expansion, based on the attached confidential case study,' try a more abstract prompt. You could ask it to 'Explain the typical economic impacts of major port expansions, citing examples from the last decade,' and then apply that general knowledge to your specific assignment on your own. This approach allows you to leverage the AI's analytical power without feeding it proprietary or personal information.
Choose University-Approved or Privacy-First Tools
Many educational institutions are now providing access to AI tools through official, licensed agreements. These enterprise versions often come with much stronger privacy protections, contractually preventing the vendor from using your data to train their public models. If your school or university offers a licensed version of an AI platform, it is almost always safer to use that than a free, public version. Additionally, a new category of privacy-focused AI is emerging. Some tools are designed to run locally on your device, meaning your data never leaves your computer, while others are built with privacy as a core feature from the ground up.
