Twelve questions to ask before trusting an AI chat app.
Privacy pages are easier to evaluate when you turn them into concrete questions. This checklist focuses on what happens to your words, files, voice, and saved memories—not vague promises about being ‘secure.’
Try it with Cappy →
1–3: Follow the prompt
Ask where inference runs, which company receives the prompt, and whether prompts may be retained or used for training. Check whether this changes by model, subscription, or account setting. A product that offers both local and hosted models should label the active path clearly at the moment you send.
4–6: Inspect storage and memory
Find out where chat history is stored, whether AI memory is opt-in, and whether individual memories can be viewed, edited, and deleted. Memory should not be a mysterious personality trait. It is stored data, and useful controls should make that visible.
7–9: Check files, voice, and web search
A text chat may be local while an attachment parser, speech service, or search feature is remote. Ask whether audio is transcribed locally, whether uploaded files are retained, and whether search queries include private conversation context. Optional capabilities should have separate, understandable boundaries.
10–12: Test exit routes
Look for a real export, a complete deletion control, and a way to clear downloaded model caches. Export files containing personal data should support encryption. Finally, verify the claims yourself: open network tools, try deleting a chat, and confirm what returns after a reload.
Quick answers
What is the first privacy setting to inspect?
Start with where inference runs and which company receives prompts. Then check retention, model training, saved memory, deletion, and whether attachments follow a different route.
Is deleting chat history enough?
Not necessarily. Saved memory, uploaded files, voice recordings, backups, and provider-side retention may be separate. A complete deletion flow should explain every storage layer.
Related guides
How On-Device AI WorksLearn what on-device AI means, what stays local, its privacy benefits, its limits, and how browser-based local AI works.User-Controlled AI Memory ControlsUnderstand AI memory, the difference between chat history and saved memory, and the controls a trustworthy assistant should provide.Local Voice Transcription and PrivacyLearn how local voice transcription works, what audio data can stay on-device, its limitations, and what to verify in a voice-enabled app.