AI Ethics Policy

Maahir is an AI-powered learning platform. We use AI responsibly and expect our users to do the same.

1. Responsible AI Use (By Maahir)

Transparency

• All AI-generated content is labelled with the agent that produced it (e.g. "✨ Generated by Roshni")
• Students are told when they are interacting with an AI versus a human teacher
• Our AI models and providers are listed publicly at /admin/ai-routes

Fairness

• We audit our AI for bias across gender, language (Urdu/English), and education level
• AI grading is reviewed by teachers for high-stakes assessments (certificates, career tracks)
• Students can appeal any AI-generated evaluation

Privacy

• AI conversations are private to the student and their teacher/admin
• We do not use student data to train commercial AI models without explicit opt-in
• AI logs are retained for 1 year then deleted

Safety

• Our AI refuses to generate harmful, illegal, or explicit content
• AI interactions are monitored for self-harm indicators; concerning conversations are escalated to a human within 4 hours
• Students under 13 receive a stricter safety-tuned AI model

2. Academic Integrity Expectations (From Students)

What's OK

• ✅ Using AI to brainstorm, get feedback, or understand a concept
• ✅ Disclosing AI assistance (e.g. "I drafted this with Roshni's help, then edited it")
• ✅ Using AI as a study buddy for practice questions and explanations
• ✅ Critically evaluating AI output as part of an assignment

What's NOT OK

• ❌ Copy-pasting AI output as your final submission without edits or disclosure
• ❌ Using AI to complete assessments designated as "no AI" (proctored exams)
• ❌ Using external AI tools (ChatGPT, etc.) when an exercise says "use only the platform AI"
• ❌ Generating fake proofs, screenshots, or deliverables with AI

3. Disclosure Standard

When submitting work that involved AI, include a brief note at the top:

> AI disclosure: I used [tool/agent] for [purpose]. I reviewed, edited, and take responsibility for the final work.

Teachers can verify any submission by requesting the AI chat log.

4. AI Detection & Appeals

• Teachers may use AI-detection tools on suspicious submissions
• A flagged submission triggers a conversation with the student (not an automatic penalty)
• Students may appeal by showing their draft history, notes, or re-completing the task live

5. Continuous Improvement

This policy is reviewed every 6 months as AI capabilities evolve. Students and teachers can suggest changes at ai-ethics@maahir.io.