AI Disclosure & Use Policy

Pathshala OS embeds artificial-intelligence (“AI”) features to help tutors and admins work faster and to give students 24/7 study help. This Policy explains where AI is used, what is sent to which provider, what AI must not be used for, and how an Institute or individual User can opt out.

• Tutor-assist — drafting lesson plans, outlines, quizzes, rubrics, and feedback comments from tutor prompts.
• Student AI chat — conversational study help limited to the topics of the User's enrolled courses; abuse and off-topic filters apply.
• Document summarisation — condensing uploaded PDFs, slides, and reading materials into bullet summaries for student preview.
• Search ranking — surfacing the most relevant lessons, materials, and Q&A across an Institute's content.
• Sales & support chat — the chatbot embedded on the public marketing site. Provides general information about Pathshala and routes complex questions to a human.

Pathshala does not use AI to grade exams, rank candidates for admission, set fees, take disciplinary action, or make any other consequential decision about a specific student.

We currently use OpenAI as our primary large-language-model provider, called from our own backend (not from the browser) so that API keys are never exposed and we can enforce content policies before/after the model call. OpenAI is listed on our Sub-processors page. We will add or change providers under the material-change rule in our Privacy Policy.

Pathshala does not use Customer Content or User prompts to train its own AI models, and does not permit its AI sub-processors to use Customer Content or User prompts to train their generic models for the public.

We use OpenAI's API tier with the “zero data retention for training” setting turned on. Prompts and completions may be retained by the provider for up to thirty (30) days for abuse-monitoring purposes only, after which they are deleted.

Pathshala does not use AI to make decisions that produce legal or similarly significant effects on a specific User (e.g., admissions, scholarships, grading, suspension). If an Institute administrator configures an automated workflow that uses AI output to take such a decision, the Institute must obtain any required consent under GDPR Art. 22, DPDP Act §10, or other applicable law; the Institute is the controller of that decision and Pathshala is the processor.

Independent of how the underlying model is prompted, Pathshala's AI features will refuse to:

• provide medical, legal, financial, mental-health, or other regulated advice as if from a licensed professional;
• generate sexual content;
• generate any sexual, violent, or harmful content involving minors (real or imagined);
• generate instructions for weapons, drugs, malware, or other illegal activity;
• impersonate a real, named person without consent;
• generate content that targets a person based on a protected characteristic.

Users will not attempt to bypass these limits (see the Acceptable Use Policy).

Every AI-generated output in Pathshala is labelled (badge, icon, or footer caption) so a User can tell it apart from human-authored content. Tutors who incorporate an AI suggestion into their own content are encouraged to review and edit it before publishing.

AI output is generated probabilistically and can be inaccurate, incomplete, or biased. Pathshala's AI features are designed to be reviewed by a human before they drive an action. Examples:

• A tutor reviews an AI-drafted quiz before assigning it.
• A student is told the AI chat can be wrong and asked to verify with a tutor or trusted source.
• A sales chatbot routes ambiguous questions to a human at hello@pathshala.co.

Institute-level. An Institute admin can turn off the AI module for the entire Institute under Dashboard → Settings → Modules. AI buttons and the student AI chat disappear immediately for everyone in the Institute.

Individual-tutor. A tutor can hide the tutor-assist surface from their own dashboard under Profile → Preferences → AI assistance. The AI chat (if their Institute enables it for students) remains available to their students.

Marketing-site visitor. Closing the AI chat widget on the public site stops further interaction; we delete the conversation transcript after 30 days.

If you encounter AI output that you believe is unsafe, discriminatory, or in violation of this Policy, please use the in-product Report button on the output (when shown) or email abuse@pathshala.co. We investigate every report and tune the safety filters based on what we find.