Governance, ethical standards, student academic integrity, and child safeguards for artificial intelligence across the Platform.
Effective date and version: October 4, 2026 (Version 2.0). This document is available in English only and forms part of, and shall be read together with, the Terms of Service and the Privacy Policy.
1.1 This Policy establishes the governance framework, ethical benchmarks, and operational rules of engagement for artificial intelligence (“AI”) across the Khypri AI platform (the “Platform”). It defines permitted and prohibited educational usage for Students, Account Holders (parents/guardians), and Educational Institutions, and expands Section 3 of the Terms of Service and Section 4 of the Privacy Policy.
1.2 This Policy governs all AI features on the Platform, including but not limited to:
2.1 Enterprise Sub-Processors: The Platform utilizes enterprise AI models supplied by vetted third-party infrastructure providers (currently OpenAI and Google Cloud Vertex AI) to perform stateless natural language inference, semantic tutoring, and safety summary drafting. All model communications take place over encrypted, server-to-server TLS connections.
2.2 Zero Legal Name Collection & Input Minimization: Inputs transmitted to inference endpoints are stripped of personal identifiers and identified solely by the Student’s parent-selected display nickname or grade tier. The Company never collects or transmits a child’s legal name, national identity numbers (such as Aadhaar), residential address, or school roll numbers to third-party AI models.
2.3 Zero Model Training Commitment: The Company relies strictly on enterprise API configurations with zero data retention and zero training commitments. Neither Khypri AI nor its backend model sub-processors use student conversations, homework submissions, audio clips, or exam responses to train, fine-tune, or reinforce foundation artificial intelligence models.
2.4 Closed-Loop “Walled Garden” Architecture: The Platform is engineered as a secure, closed-loop educational sanctuary. The AI tutoring agents (Olli and Lumi) are programmatically prohibited from generating, outputting, or embedding external hyperlinks, web search engine redirects, or third-party web URLs into student conversations. Students are never redirected to unverified external internet destinations or open web search environments.
2.5 Zero Third-Party AI Referrals: The Platform strictly prohibits the AI from referencing, recommending, or providing access links to external third-party AI platforms (such as Google AI, Gemini, ChatGPT, or external conversational bots). All educational interactions remain completely self-contained within Khypri, preventing exposure to external commercial platforms, cookies, or unmonitored tracking.
3.1 Socratic Learning Philosophy: Khypri AI is designed to foster critical thinking, deep conceptual mastery, and active recall. The AI operates on the Socratic method—guiding students through step-by-step reasoning rather than simply providing instant answers.
3.2 Permitted AI Study Uses: Students are encouraged and authorized to utilize the AI Platform for:
3.3 Prohibited Conduct (Academic Dishonesty & Fraud): In alignment with CBSE, ICSE, and global academic integrity directives, the following acts are strictly prohibited and constitute a material breach of the Acceptable Use Policy:
4.1 Non-Human Nature Disclosure (MeitY AI Advisory 2024–2026): Pursuant to regulatory advisories issued by the Ministry of Electronics and Information Technology (MeitY), the Platform explicitly identifies the artificial intelligence nature of its tutoring agents in all chat interfaces. Students are continually reminded that Olli is an automated synthetic study companion, not a human teacher, peer, or counselor.
4.2 Parasocial Attachment Guardrails: To protect adolescent psychological development, the AI is engineered with strict emotional boundaries. It is prohibited from simulating romantic affection, reciprocating pseudo-intimacy, or isolating a student from parental and teacher relationships. When a student expresses loneliness, grief, or personal despair, the AI immediately de-escalates and urges speaking with a parent, school counselor, or trusted adult.
4.3 Synthetic Media Watermarking: In compliance with national digital safety frameworks, all AI-generated practice materials, synthesized voice audio, and interactive worksheets carry cryptographic metadata watermarks identifying them as synthetically generated educational content.
5.1 Inherent Hallucination Risks: Generative AI models operate on probabilistic language patterns and can occasionally generate inaccurate, incomplete, or outdated responses (“hallucinations”). AI output is a supplementary study aid and must never be treated as authoritative legal, medical, or definitive factual truth.
5.2 Official Curriculum Textbook Primacy: In any circumstance where an explanation or formula provided by the AI diverges from prescribed national or state board textbooks (including NCERT, CBSE, ICSE, Cambridge, or State Board publications), the official prescribed textbook shall strictly govern and supersede the AI output for all school evaluations and board examinations.
5.3 Algorithmic Non-Discrimination (Article 14, Constitution of India): In accordance with constitutional principles of equality under Article 14 of the Constitution of India, the Company continuously audits prompt architectures and model weights to eliminate algorithmic bias across gender, religion, caste, language, regional origin, or socio-economic background.
5.4 Intellectual Property & Fair Dealing: Practice questions generated on the Platform are provided for personal family study under Section 52(1)(a) Fair Dealing provisions of the Copyright Act, 1957. Proprietary prompt configurations, system rules, and avatar identities remain the exclusive intellectual property of the Company.
5.5 Operational Rate Limits, Evaluation Scope & Infrastructure Governance: To prevent computational resource exhaustion, mitigate automated denial-of-service degradation, and guarantee rapid inference latencies for all active students, the Platform enforces automated velocity throttles and evaluation quotas: (a) Olli: 10 queries/minute and 250 queries/hour; (b) Lumi: 250 MCQ practice marks/hour; (c) Axis Mock Papers: 500 marks/24 hours; and (d) Axis Subjective Evaluation: 1 lifetime evaluation per Account for Free Tier, and up to 10 evaluations per rolling 30-day period per Account for Paid Subscriptions (shared across all registered student profiles). Complete quota criteria, non-rollover terms, and telemetry verification rules are governed under Section 7.5 of the Terms of Service.
6.1 Human-in-the-Loop Teacher Verification: For schools and educators utilizing the Axis engine to draft mock papers, the Platform enforces a mandatory teacher-in-the-loop review. Educators must review question clarity, syllabus alignment, and marking rubrics before publishing or printing examination papers for students.
6.2 Pedagogical Rubric Alignment: Exam generation algorithms enforce strict pedagogical compliance with Bloom’s Taxonomy cognitive levels (Remembering, Understanding, Applying, Analyzing, Evaluating, Creating) and official board blueprint weighting.
7.1 Continuous Safety Screening: Student exchanges with the AI are analyzed in real-time by Khypri’s multi-tier Safety Core™, classifying interactions across 158 normalized child safety and risk categories, ranging from self-harm and bullying to substance abuse and campus security.
7.2 Parental Notification & Crisis Escalation: Where a high-priority safety risk is detected (such as acute depression or distress), the AI immediately pivots to a supportive crisis de-escalation posture, highlights statutory emergency helplines (Tele-MANAS: 14416 / 1800-599-0019, National Emergency: 112, POCSO: 1098), and transmits high-priority notifications to the Account Holder.
7.3 Full Account Holder Transparency: Account Holders retain unrestricted access to their child’s learning history, session transcripts, diagnostic progress, and safety telemetry through the Parent Portal.
8.1 Inclusive Interface Standards: In accordance with Sections 40 and 46 of the Rights of Persons with Disabilities Act, 2016 (RPWD Act) and WCAG 2.1 AA accessibility guidelines, the AI interface provides full screen-reader semantic labeling, high-contrast visual themes, keyboard navigability, and multi-language support.
9.1 Policy Enforcement: Violations of this Policy may result in prompt throttling, temporary feature locking, or account termination. Any attempt to exploit AI systems to harm children or compromise national infrastructure will be reported to appropriate law enforcement authorities pursuant to the Information Technology Act, 2000 and Bharatiya Nyaya Sanhita, 2023.
9.2 Grievance Redressal & DPO Inquiries: Inquiries, algorithmic bias concerns, or feedback regarding AI safety should be directed to: