We want AI to make students think harder, not less.

We are engineering students. We have watched classmates paste an assignment into a chatbot and submit whatever comes back. They pass the course, but they skip the learning the assignment was there for.

Teachers can't see any of this. Banning AI doesn't work, because every student has a phone. AI-text detectors are unreliable and accuse honest students, so teachers stop trusting them.

Meanwhile fast open models have become cheap enough that a personal tutor for every student is affordable for an Indian college for the first time.

So we are building the tutor we wish we had: Sage coaches with hints and questions instead of handing over answers, and follows the rule the teacher sets for each assignment. Lens answers a teacher's questions about their class in plain English. Trace shows how a piece of writing was made, so a conversation can start from evidence instead of suspicion.

Most products in this space sell monitoring and filtering to schools, on devices the school controls. They watch students.

SafeSight is built for Indian colleges and coaching institutes, and it works the other way round. Students use Sage because it helps: it knows their course notes, the assignment and where they are stuck. The teacher's view comes from that use, not from surveillance.

• Teachers see summaries and scores by default. Full chats open only for flagged items, or when a student shares one.
• Live Lab Mode runs only during announced exams on college lab PCs, and switches off when the session ends.
• Student data is never used to train AI models.

Read the details in plain English in our privacy policy.