Thinking & Safe Use · 4 MIN READ
Protecting sensitive information
Decide what belongs in an AI tool before sharing it.
Pause before you paste
Check your organisation’s rules and the service settings before sharing work material. Permission to read a document does not automatically mean permission to upload it to another service. These are practical handling principles; requirements depend on your setting.
- Look for personal details, client information, credentials and unpublished work.
- Use an approved service and share only what the task requires.
- If permission is unclear, use a fictional example or ask the information owner.
Remove detail that the task does not need
Replace names with role labels, omit account details and use invented values when the exact figures are unnecessary. Removing a name alone may not prevent identification: combinations of location, role and events can still reveal someone.
Original task: improve the structure of a client update. Safer input: a fictional update with the same headings and tone, without client names, contract terms or contact details. Request: comment on structure only; do not infer missing facts.
Check the output before sharing it
Review generated work for details copied from the input or invented by the model. An anonymised input does not make an output automatically appropriate for publication. If sensitive material was shared accidentally, follow your organisation’s incident process.
Explore the webinar for guided practice, or subscribe for new perspectives.
Explore The Webinar ↗Subscribe Newsletter ↗