LAST UPDATED: September 10, 2026 • JURISDICTION: INDIA • VERSION 2.5

AI Services & Responsible AI Policy

Overview: Consisto integrates state-of-the-art artificial intelligence services (collectively "cAI" and "CAL" — Consistent AI Lab) powered by industry-leading foundation models to assist students and teachers in interactive learning, assessment, automated paper creation, and community moderation. This Policy outlines supported AI models, data privacy guarantees, non-training commitments, and pedagogical boundaries.

1. AI Providers & Models Used

Consisto cAI services process queries dynamically using enterprise AI foundation models to ensure high availability, speed, and accuracy:

  • Groq Models: openai/gpt-oss-20b (primary high-speed inference engine for real-time tutoring and paper creation).
  • Google Gemini Models: gemini-3.1-flash-lite (high-capacity context engine for subjective evaluation and batch grading).
  • DeepSeek Models: deepseek-chat (standby academic inference engine).

2. Student Assistance & Socratic Educational Boundaries

Inside the Consistent AI Lab (CAL), cAI acts as an academic tutor for students across NCERT, CBSE, JEE, and NEET curricula:

  • Socratic Guidance Methodology: cAI provides step-by-step conceptual explanations and LaTeX mathematical formatting ($...\$) rather than direct answers, fostering active problem-solving skills.
  • Personalized Practice (Adaptive DPPs): cAI generates customized Daily Practice Papers tailored to student accuracy metrics and learning progress.
  • Non-Replacement of Instruction: cAI is a supplementary learning tool. It does not replace formal classroom instruction, accredited textbooks, or human teacher guidance.

3. Teacher Tools & Assessment Assistance

For educators, cAI automates time-consuming administrative workflows while preserving human pedagogical oversight:

  • Automated Exam & Test Creation: Teachers can generate grouped question sections (MCQ, Short Answer, Long Answer, Assertion & Reason) aligned with NCERT difficulty distributions.
  • CBSE Subjective Step-Marking Evaluation: cAI evaluates student free-text answers against official marking schemes and teacher rubrics. Final grading authority remains 100% with the human teacher.
  • Classroom Copilot Insights: Aggregates performance trends to highlight class-wide knowledge gaps and suggest revision plans.

4. Strict Non-Training Data Privacy Guarantee

Consisto enforces non-negotiable data isolation and privacy controls for all AI requests:

🔒 Zero Public Model Retraining Commitment: Student queries, submitted answers, test papers, and academic records sent to cAI provider APIs are processed exclusively for real-time inference. Your data is NEVER used to train, fine-tune, or improve public AI models operated by Consisto, Groq, Google, DeepSeek, or any third party.
  • Data Minimization: API requests strip personal identifiers before sending academic prompts to subprocessor APIs.
  • Encrypted Transient Processing: Prompts are processed ephemerally in server memory and discarded immediately following response delivery.

5. Automated Content Moderation (cAI-cMS)

The cAI Community Moderation System (cAI-cMS) inspects student community posts and public discussion boards in real-time. Operating alongside dictionary filters, cAI-cMS detects harassment, cyberbullying, profanity shortforms, and safety threats, taking instant protective action (warning, 24h mute, or flag for review).

6. Limitations & Disclaimer of AI Accuracy

While cAI is tuned against NCERT and CBSE academic benchmarks, generative AI technologies may occasionally produce incomplete or inaccurate content ("hallucinations").

  • Users should independently verify critical mathematical proofs, historical dates, or scientific constants.
  • Consisto disclaims liability for academic score variations or decisions made based upon AI-generated outputs.