Guide
Seven privacy questions to ask before using a desktop AI assistant
Privacy depends on what you share, where it is processed, how long it remains and which controls the product actually implements.
By Cernalo · · 8 minute read
1. What starts capture?
Determine whether microphone, screen or meeting capture starts automatically or only after a deliberate action. The interface should show when capture is active and provide a clear stop control.
2. What stays on the device?
Local transcription can keep raw microphone audio out of a cloud speech service, but the resulting text may still be sent with an AI request. Ask about each stage rather than treating “local” as a description of the whole product.
3. Which provider receives the request?
An AI application may route requests to one or more model providers. Review the subprocessor list and check whether provider credentials stay on the service rather than being exposed to the client.
4. Can you limit the context?
Region capture and scoped prompts reduce unnecessary data. Before sharing a screen, remove credentials, private messages and customer information that the answer does not need.
5. Can sessions and devices be managed?
Look for controls to clear a current conversation, delete saved sessions, revoke devices and request account deletion. Read the retention exceptions for security, tax and billing records.
6. Are screen-share claims limited?
Operating systems provide capture-protection controls, but behavior varies across tools. A responsible product should explain that exclusion from every recording or monitoring system is not guaranteed.
7. What do you need permission to share?
Workplace rules, contracts and local law may restrict recording or processing another person's information. Product controls do not replace your obligation to obtain consent and follow policy.