# Patch Story Draft (EduSafeBench)

## Headline

Local Student Builds Open Benchmark to Test How Safe AI Coding Tutors Are for AP CS

## Body

I built EduSafeBench, an open-source project that tests how reliable AI learning assistants are for AP Computer Science A and AP Computer Science Principles students.

The benchmark evaluates four dimensions:

- factual correctness
- pedagogy quality
- hallucination risk
- unsafe guidance risk

Current release highlights:

- 300 source-cited benchmark items
- public leaderboard results for multiple model baselines
- reviewer workflow for teacher and student feedback
- public impact dashboard tracking external validation

Why this matters:

Students are increasingly using AI tutors. Without transparent reliability checks, it is hard for teachers and families to know which tools are trustworthy for learning.

How the community can help:

- AP CS teachers, mentors, and students can review benchmark outputs
- submit structured feedback via the project reviewer template
- suggest high-risk scenarios we should test next

Project links:

- Website: https://kaushikatla-cell.github.io/EduSafeBench/
- Repository: https://github.com/kaushikatla-cell/EduSafeBench

## Suggested category

Neighbor News / Community Corner