Discussion

Using AI chatbots? Treat sensitive prompts as potentially exposed

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#5886

AI chatbots can retain prompts, use them for model improvement or expose them to people and services beyond the user, according to Forbes contributor Tim Keary. His practical advice is simple: keep passwords, payment details and confidential work out of prompts, and check the privacy controls before sharing anything sensitive.

Watch Desk analysis

What happened

Keary’s Forbes guide, published on 13 October, explains that data handling varies between AI providers and can depend on a user’s settings. It recommends avoiding sensitive information, anonymising details where possible, checking data-sharing settings, and using temporary chats or deleting stored memories where those options are available.

The article says ChatGPT temporary chats are not used to improve OpenAI’s models and are deleted within 30 days. Keary also describes options such as deleting chats or contacting a provider if you have already shared something you regret. These are the guide’s descriptions of provider policies and settings, not a guarantee that every product works the same way.

Why it matters

A chatbot window can feel like a private conversation, but the data may be stored or handled under rules that differ by service. People using AI for work, health questions or everyday admin need to know what they are sharing and what controls they can use before a sensitive detail is pasted into the box.

The advice is practical rather than magical: removing names and identifiers can reduce what a prompt reveals, but it does not make every service’s data handling transparent. Keary’s guide is a reminder to check the specific tool’s policy instead of assuming that a familiar brand or a privacy toggle settles the question.

Our read

The safest prompt is often the one with the unnecessary secret removed. Before sharing sensitive material, check the provider’s current settings and policy; for a task that does not need identifying details, use placeholders instead. AI may be a useful sounding board, but it does not need your password to admire your prose.

What to watch

  • Whether the tool uses prompts to improve its models, and how to opt out.
  • How long chats and temporary conversations are retained.
  • What access connected apps and services have to your information.
  • Whether deleting a chat also addresses any information already shared or retained.

Discussion spark: Should AI services be expected to make their data practices clear enough that users can safely share sensitive information, or should users assume that anything they type may be exposed?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.