As artificial intelligence chatbots such as ChatGPT, Copilot, and Gemini become integral to everyday tasks, from summarising documents to generating creative ideas, many users are unaware of how their personal data is handled. While these services offer convenience, the information shared – including prompts, conversations, attachments, and even location data – can often be stored by providers, raising significant privacy concerns for UK consumers and businesses.
Companies typically state that storing user interactions helps them refine and improve their AI models. However, this practice means that what feels like a private conversation with an AI assistant is, in reality, an interaction with an online service where data is retained. Experts advise that before sharing sensitive details or confidential files, users should consider whether they would be comfortable with that information being permanently stored in an online account or cloud application.
Fortunately, individuals can take proactive steps to regain control over their digital footprint within these AI tools. Simple adjustments to privacy settings within the apps can significantly limit the scope of data collected and retained. For instance, in ChatGPT, users can navigate to 'Data Controls' to switch off the option that allows chats, prompts, and attachments to be used for improving OpenAI's models. There are also options to delete previous conversation histories and manage or delete 'memory' features, which store ongoing context about user preferences.
Similarly, Google's Gemini allows users to control their 'Gemini Apps Activity'. By turning this off, prompts and conversations will not be stored in their Google Account. Users can also set up auto-delete functions, enabling saved activity to be automatically removed after specified periods, such as three, 18, or 36 months. For those seeking maximum privacy, many chatbots offer 'temporary chat' features, where conversations are not saved and memory functions are disabled.
Beyond adjusting settings, the most secure approach remains a mindful interaction with AI tools. Users should only share the minimum amount of information genuinely required for a task. If there's any doubt about the sensitivity of data, it is advisable to redact personal details from documents or upload anonymised versions rather than the originals. This cautious approach helps mitigate potential risks associated with data retention and usage by AI providers.