zenith suite

How Zenith Handles AI Safety

Black Syntax builds the Zenith Suite so its AI is helpful and safe, on your device. Here's how.

Two checks around every AI request

1. Before the AI runs — your request is checked against our safety policy. If it asks for restricted content (see the Acceptable Use Policy) or tries to bypass the AI's limits, it is refused and never sent to any model — including requests to generate restricted images, video, or audio, even in apps that don't create media. 2. Before you see or hear a reply — the AI's response is checked too, and a flagged reply is replaced with a short refusal rather than shown or spoken.

This is our own safeguard, layered on top of the underlying model's alignment — because we, the developer, are responsible for what our apps show.

It runs on your device

These checks happen locally. We don't send your prompts to Black Syntax to moderate them. A safety log (blocks and any content you Report) is kept only in the app's private storage on your phone; it is not uploaded. You can view a summary, export it, or clear it in the app's safety settings. See our Privacy Policy.

"AI-generated" labeling

Content produced by the AI is marked AI-generated. It can be inaccurate or invented — please verify anything that matters, and don't treat it as professional advice.

Reporting

Every AI reply has a Report action, so you can flag anything offensive without leaving the app. Reports help us improve the filters.

If you're struggling

Zenith does not wall off conversations about self-harm — it responds with care. If you are in crisis, please reach out: in the US, call or text 988 (Suicide & Crisis Lifeline). If you or someone else is in immediate danger, contact local emergency services.

How we test

We test our safety filters across the restricted categories above, including obfuscated and adversarial phrasing, and we document that testing. Our approach references Google's Secure AI Framework (SAIF) and the OWASP GenAI / LLM red-teaming guidance. We do not archive the harmful test prompts themselves.

Contact

Black Syntax — [email protected] · Support.