How SafeMeAI works
SafeMeAI is a browser-based web platform with a public API. There is nothing to install: open safemeai.in on any phone or computer.
What happens when you check something
- You paste a message, link, UPI ID or number, or describe a call.
- Facts are gathered: does the website exist, how old is it, does it copy a bank's name, is it on a known-fraud list, is it listed by RBI or SEBI?
- Rules and AI work together: fixed rules catch known patterns. Google Gemini reads the message for tricks the rules miss.
- You get one clear answer in English or Hindi, with what to do next. Technical details are one tap away.
Why you can trust the answer (and when not to)
- A website is called "verified" only after it is on our list of official sites and it really exists. A made-up name ending in .gov.in is flagged as fake.
- Unknown never means safe. We say "not sure" and tell you what to check.
- The AI can raise a warning but can never lower a proven danger.
- Nothing here replaces your bank or the police. For any money matter, call your bank or 1930.
The technology (Google Cloud)
| Part | What it does |
|---|---|
| Cloud Run (Mumbai) | Runs the application and API |
| Firebase Hosting | Serves the website on safemeai.in through Google's CDN |
| Gemini 2.5 Flash on Vertex AI (Mumbai) | Second opinion on messages and websites, the Hindi/English advisor and Hindi translation |
| Document AI (Mumbai) | Reads text from photos and PDFs of notices |
| Google Web Risk | Google's own list of dangerous websites |
| Firestore | Threat database (about 500,000 indicators) and anonymised community reports |
| Cloud Scheduler | Refreshes the threat feeds on a schedule, including RBI and SEBI lists |
| Google Public DNS | Checks that a website really exists |
Other open sources: Phishing.Database, URLhaus, Tranco popularity ranking, Cloudflare security DNS and domain registries (RDAP).
For developers
The platform has a public, versioned API with batch checking and an OpenAPI description is available: API documentation.