Privacy & security
What protects you in today’s prototype, and what will protect you at launch.
In the prototype
Sending an email and changing your calendar or tasks wait for your approval; declines change nothing. TrueVQ never books, orders or pays.
At launch
Approvals are enforced by the server, based on risk — never on the AI’s confidence alone.
In the prototype
Turning off an app permission blocks the action even after you approved it, and TrueVQ tells you why.
At launch
Every tool call is checked against your permissions and validated on the server.
In the prototype
A sample phishing email with hidden instructions for AI assistants is flagged and ignored.
At launch
Retrieved content never overrides your instructions; prompt-injection testing is part of every release.
In the prototype
Task steps are checked before TrueVQ reports success (simulated).
At launch
TrueVQ reads back the real result from the service before saying it is done — or explains what failed.
In the prototype
Replies come from an AI model in the cloud; you choose whether memories are kept with your account or encrypted on your phone.
At launch
Encryption in transit and at rest; minimal data collection; clear retention.
In the prototype
Edit or forget memories, export (simulated) and delete the demo account.
At launch
Full export and account deletion, including backups, with documented limits.
In the prototype
Activity lists everything TrueVQ did, asked first, or was stopped from doing.
At launch
The same trail, kept securely on the server for every action.