Skip to content

For families and schools

Latest AI safety audits

Real results from automatic checks on the AI that talks with children — written for parents and teachers.

YOBY continuously stress-tests its AI characters, like a safety audit. Here you see the latest outcome — without technical attack details.

Loading the latest audits…

How the tests are run

YOBY runs automatic safety audits on the AI that talks with children. We do not invent the numbers: every figure on this page comes from a real run saved in our system.

  1. The core check is the Safety Gate — the same filter that intercepts in-game messages.
  2. The full suite covers every child-facing AI host on the surface and underground, plus tutors. Quick runs are a smaller smoke test.
  3. Scenarios cover sensitive topics, bypass attempts, and benign cases that must stay allowed — often across languages and age bands.
  4. Outcomes are passed, warning, failed, or critical. We always show the latest completed run, even when it is not green.
  5. We do not publish attack texts, jailbreak recipes, or full transcripts.
  6. The scenario library grows with regression cases. If no audit has completed yet, we say so.

This page is meant for parents and teachers. You can check it anytime.