Using AI to help people memorize the Quran is a serious responsibility. Here is how we handle it.
Guiding principles
- The Quran is not AI-generated. Every verse, transliteration, translation, and tafsir on the app comes from authenticated sources, not from generative models. AI is used for scheduling, scoring, coaching, and explanation — never for producing the Quran itself.
- Authority belongs to scholars, not algorithms. The AI Quran Coach assists with memorization technique and discipline. For religious rulings, it points you to our Scholar Panel or qualified external scholars.
- Your voice is yours. We analyze recitation audio to verify pronunciation. We do not use your audio to train our own or any third-party models. Audio is deleted from our servers within 30 days.
- Transparency. Every AI-generated response is marked with an “AI-generated” label. The model and version are disclosed.
- Human oversight. All religious content is reviewed by the Scholar Panel before publishing.
Which AI models we use
| Purpose | Primary model | Fallback |
|---|---|---|
| AI Quran Coach chat | Google Gemini 1.5 Flash | DeepSeek |
| Speech transcription | Native iOS / Android speech APIs | Cloud STT |
| Phonetic verification | In-house matcher | — |
| Recommendation engine | In-house gradient-boosted model | — |
We publish this list and keep it updated in our AI Safety Changelog.
Safety guardrails
- The AI Coach will refuse to issue religious rulings.
- It will refuse to answer questions outside memorization, hifz discipline, and well-established Islamic learning topics.
- It is red-teamed monthly against common adversarial prompts (fabricated verses, disputed rulings, identity spoofing, emotional distress escalation).
- A human-escalation path is one tap away: “Ask a scholar.”
Your controls
- Disable AI Coach. Settings → AI → Disable.
- Delete recitation history. Settings → Privacy → Recitation audio → Delete now.
- Opt out of cloud analysis. Settings → AI → On-device only.
Reporting an issue
If the AI Coach says something wrong or inappropriate, tap “Report” inside the chat, or email ethics@hafizprime.com. We acknowledge within 48 hours.
