~ Inspiration: We built this because of something that happened in our own homes. Last month, our grandparents received a terrifying SMS: "Your electricity will be cut off tonight at 9:30 PM. Pay immediately."They panicked. Millions of parents and seniors face this every day, losing their life savings to digital fraud. Existing spam blockers only block numbers, but scammers just buy a new ₹50 SIM card and target someone else. We thought: "Instead of letting our families be defenseless targets, why can't we give them a digital bodyguard that turns the tables and traps the scammers?" That’s how"ScamBuster AI" was born. ~ What it does: ScamBuster AI is an autonomous digital bodyguard that protects citizens and catches fraudsters.
- Zero New Apps to Learn: When an elderly person gets a scary SMS or WhatsApp message, they simply "forward it or send a voice note" to our ScamBuster WhatsApp contact.
- Instant Peace of Mind: Within 2 seconds, our AI reassures them in their own mother tongue (Hindi, Kannada, etc.): "Don't worry, this is a scam. Your power is safe. We are handling it."
- The Autonomous Honeypot: From a separate, safe virtual burner number, our AI agent steps in and chats with the scammer. It plays the role of a confused, polite elder, keeps the scammer busy, and "tricks them into sharing their real UPI ID and bank account number".
- 1-Click Police Report: The AI automatically packages the chat logs and extracted bank details into a ready-to-submit "Cybercrime Police Complaint (PDF)". ~ How we built it: We kept our tech stack focused on speed, safety, and simplicity:
- The Brain: We used the "Google Gemini API"for fast reasoning and natural conversations in regional Indian languages.
- Agent Logic: We used "LangGraph"to manage the workflow: check message and to calm user to bait scammer to extract bank details to create PDF.
- Safety & Decoy Lines: We used cloud virtual numbers so the scammer "never sees the real user’s phone number", and an isolated sandbox browser to inspect scam links safely.
- Frontend & Demo: A clean "Streamlit dashboard"that shows the live chat and auto-generates the final police report. ~ Challenges we ran into:
- Respecting WhatsApp Privacy: We didn't want to compromise WhatsApp's encryption. We solved this by using a simple "Forward to Contact" model, so user privacy is 100% protected.
- Making the AI Sound Real: Scammers know it's a bot if replies are instant and robotic. We had to program human typing pauses, slight hesitations, and everyday conversational language so scammers believe they are talking to a real person.
- Avoiding False Alarms:We made sure our system can easily tell the difference between a real bank alert and a fake scam message before releasing the agent. ~ Accomplishments that we're proud of:
- Built for our Families: We created something our own grandparents can use without having to learn anything new.
- Real Evidence Gathering: Seeing our AI successfully trick a scammer into giving their active UPI ID within just 3-4 chat turns was an incredible moment for our team.
- Turning the Tables: We took power back from scammers and put it into the hands of ordinary citizens. ~ What we learned:
- Building an autonomous agent isn't just about writing prompts—it’s about teaching the AI to plan, take safe actions, and evaluate its own work.
- How scammers psychologically exploit fear and urgency.
- The best technology is the one designed with deep empathy for the people using it. ~ What's next for ScamBuster AI:
- Direct Police Helpline Link: Partnering with cybercrime units (like the 1930 portal) to automatically freeze extracted bank accounts in real time.
- Voice-Call Answering: Giving ScamBuster a synthetic human voice so it can answer actual scam phone calls and waste fraudsters' time on the phone.
- More Local Languages: Expanding our vernacular support to protect citizens across every corner of India.
