AI Use Policy

1. How AI Is Used

• AI may help summarize uploaded notes, identify concepts, draft private study packs, explain mistakes, generate practice suggestions, and support parent-facing summaries.
• Question-bank generation is handled as a governed supply chain. Candidate content must pass validation and admin curation before it becomes reusable student-visible inventory.
• Scored or strict-test student sessions should serve deterministic, approved content rather than waiting on live AI question generation.

2. Practice Mode

Practice mode may include hints, explanations, formula references, worked examples, note-derived study packs, and other support tools. These supports are for learning and may be limited, collapsed, or logged so the student experience stays focused.

3. Strict Tests

Strict tests should not provide AI tutor help, hints, scaffolds, solution steps, explanation views, code-runner exploration, or other answer-assistance tools during the test unless TARK has explicitly configured an approved accommodation or official exam-authorized resource. Official resources, such as an exam reference sheet, are separate from AI assistance.

4. Student Data and Private Uploads

AI requests may use uploaded notes, extracted text, answer context, topic signals, mistake patterns, or mastery signals when needed to provide the requested learning feature. Students and parents should not upload medical records, financial records, government IDs, unrelated family documents, or other sensitive information that is not needed for study.

5. External LLM Boundary

• External LLMs may be used manually for generic curriculum drafts or generic question drafts only when no real student, parent, learner-code, PIN, upload, answer, or profile data is included.
• Student-specific generation should use TARK's approved internal processing path, not a pasted prompt to a consumer chatbot.
• External drafts imported into TARK Learn are inactive candidates. They are not student-visible until they pass validation and admin curation.

6. AI Can Be Wrong

AI output can be incomplete, ambiguous, outdated, or incorrect. TARK Learn uses validation, curation, and student feedback to reduce this risk, but parents and students should report confusing questions, wrong answer keys, copied material, or unsafe output.

7. Contact

Questions or requests can be sent to privacy@tarklearn.com. Product support can be sent to support@tarklearn.com.