> For the complete documentation index, see [llms.txt](https://whitepaper.huralya.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://whitepaper.huralya.com/safety-privacy-and-trust.md).

# Safety, privacy and trust

A companion people open up to must earn their trust. Hura builds safety into the system itself, not into a line of small print.

* **Care that scales with risk.** Every reply is checked by two independent layers, the AI's own judgement and a deterministic language net. The stronger signal always wins. When there is real danger, Hura responds with clear support and local emergency numbers; ordinary sadness is met with warmth, not alarms.
* **Built for many languages.** Risk phrases are recognised in English, Persian, Arabic, Turkish, Spanish and Uzbek, including everyday spoken forms.
* **Honest about its limits.** Hura is a wellbeing companion. It does not diagnose, it presents self-check results as information, and it is open that it does not replace therapy or medical care.

Privacy follows the same principle: collect less, protect more.

* **Start anonymously.** No email or phone number is needed. A private recovery phrase keeps the account safe, and sign-in with Google, Apple or Telegram is optional.
* **Your data stays yours.** No ads, no data sales, no sharing with third parties. Journal and habits live on the device first.
* **Secure by default.** All traffic is encrypted, and deleting an account erases it completely, including the AI's memory.
