SaveYours AI

Inspiration

Children can face online risks that develop gradually through repeated interactions rather than a single suspicious message. We wanted to build a tool that helps people recognize concerning patterns, understand potential risks, and respond supportively without invading privacy or making automatic accusations. This inspired SaveYours AI.

What it does

SaveYours AI is a privacy-conscious child online safety prototype that analyzes conversations to identify potentially concerning signals, such as secrecy, manipulation, pressure, and boundary violations. It provides a concern level, explains the signals behind its assessment, and suggests supportive next steps. The goal is to assist human judgment, not replace it.

How we built it

We built SaveYours AI using React, Vite, JavaScript, Node.js, and Express.js. The frontend provides an interactive interface for analyzing conversations and reviewing results, while the backend processes requests through a safety analysis engine. The current baseline uses transparent, rule-based detection and synthetic test scenarios to demonstrate evaluation. AI-assisted analysis is a planned or optional extension where implemented and configured.

Challenges we ran into

One of our biggest challenges was identifying meaningful patterns without treating every unusual message as dangerous. We also had to consider false positives, missed signals, privacy risks, and the limitations of simple keyword-based detection. Designing clear explanations while avoiding unsupported accusations was another important challenge.

Accomplishments that we're proud of

We developed a working prototype that demonstrates conversation analysis, interpretable risk signals, concern-level assessment, and supportive safety guidance. We are particularly proud of prioritizing privacy, transparency, and responsible human review instead of relying on unexplained scores or automatic escalation.

What we learned

We learned that child safety requires more than detecting keywords. Context, repeated behavior, privacy, and the possibility of false alarms all matter. We also learned the importance of testing with synthetic scenarios, communicating limitations honestly, and designing technology that supports people rather than making decisions for them.

What's next for SaveYours AI

Next, we aim to improve contextual understanding, evaluate more diverse synthetic scenarios, reduce false positives and missed detections, and strengthen privacy protections. We also plan to validate AI-assisted analysis against the rule-based baseline, improve multilingual support, and conduct careful testing with appropriate safeguards before considering real-world deployment. Our long-term vision is to make SaveYours AI a reliable, explainable, and privacy-conscious tool for supporting children's online safety.

Built With

• css
• engine
• express.js
• gemini
• javascript
• node.js
• react
• restapi
• vite