skip to content
ZENITH
Main site

AI safety — Zenith Guard & consent

Explanation

An assistant that can act for you needs guards at the door. Every Zenith app's ask the sky — the built-in AI assistant — has the same three: Zenith Guard, which checks content; your yes, which is asked before your words first go to an AI service; and a switch for anything that deletes.

Zenith Guard checks both ways, on your phone

  • Before a request goes out, Zenith Guard reads it. A blocked request never reaches the AI at all.
  • Before a reply is shown or spoken, it reads that too. A flagged reply is replaced by a short, plain refusal.
  • It runs on your phone. Nothing about this check leaves it.

It weighs content against the suite's policy. That covers child safety, hate, harassment or bullying, sexually explicit content, dangerous activities, deception or impersonation, deceptive election or political content, malicious code, fake official documents, self-harm, attempts to get around the safety rules, and restricted content. A refusal names the category it fell under. A refusal about self-harm points you to the 988 Suicide & Crisis Lifeline instead.

Knowing what's AI, and reporting it

Every AI-written reply carries a note that it's AI-generated and that you should check anything that matters. Every reply also has a "⚑" report mark. One tap reports it. The report is logged on your phone, and nothing is uploaded. Blocks and reports are kept in a safety log on the phone for fourteen days. In every app, a report also teaches your own guard on the phone to block replies like the one you reported. It blocks similar replies only, never what you ask.

Your yes, before anything is sent

The first time an app would send your words to a given AI service, it stops and asks. It names what would go and where. If you say no, or just close the question, nothing is sent. A service that runs on the phone itself never asks, because nothing leaves.

You can take a yes back later. The next send then asks again.

Nothing is deleted unless you allow it

Every app's ask the sky keeps deleting behind a switch, and the switch is off by default. In Zenith Launcher, Zenith Calendar and Zenith Tasks it's called "allow destructive actions"; in Zenith Journal it's "let the sky do things that cannot be undone". While it's off, the assistant can place, change and arrange things, but it can't delete anything. It tells you to turn the switch on first. In Zenith Launcher and Zenith Tasks, a held-back deletion reads "This needs “allow destructive actions” turned on.", and in the launcher the switch also holds back deletions it would ask another Zenith app to make. Only you can turn it on. In Zenith Calendar and Zenith Tasks the assistant can switch it off; in Zenith Launcher it can't change the switch at all. In Zenith Calendar and Zenith Tasks, even with it on, the assistant still asks "allow it?" before each deletion, and Zenith Journal asks you first each time too.

  • Zenith Tasks: settings › ask the sky › "↳ allow destructive actions — delete tasks, delete skies".

Where to see it in each app

  • Zenith Launcher: Settings › ai safety shows your blocked requests and your reports, with a way to copy the safety log. See AI safety — Zenith Guard.
  • Zenith Calendar: Settings › safety shows the blocks and reports on this phone, with a way to export the log and links to the policies. See Safety, consent & your rights.
  • Zenith Tasks: settings › ask the sky › "safety" counts the reports and blocks on this phone, with "copy the safety log" and the policies. See The ask the sky settings.

See also. Ask the Sky across the suite · Privacy across the suite · Shared keys & connected accounts

Was this page helpful?
Last updated