Gmail AI Threat Scanner: An Open‑Source Security Layer for Your Inbox

Email remains the most common entry point for cyberattacks, yet most people rely entirely on whatever built‑in filtering their provider offers. I wanted something more transparent, more customizable, and more aligned with modern AI‑driven threat analysis. So I built Gmail AI Threat Scanner.

This project connects to your Gmail inbox, analyzes messages using AI‑powered threat detection, and flags suspicious content — all while running locally so you stay in control.

What it does
Scans Gmail messages using AI models for phishing, fraud, and social engineering indicators

Highlights risky emails with clear explanations

Works entirely client‑side using OAuth (no passwords, no server storage)

Lets you review threats in a simple, readable interface

Why I built it
I wanted a lightweight, open‑source tool that gives users visibility into potential threats without requiring enterprise‑grade infrastructure. Security shouldn’t be locked behind corporate paywalls — especially when AI can help individuals defend themselves.

Roadmap

• Add multi‑model ensemble scoring
• Add attachment scanning
• Learning from False Positives

GitHub Repo
🔗 https://github.com/ajaynyayapathi/Gmail-AI-Threat-Scanner

If you try it, I’d love feedback, issues, or PRs.

Comments

[Furqan Ashraf]

Cool project. Since you asked for feedback: right now every verdict comes from the model, so a deterministic check before the AI call would help. Pull the sender domain and every link domain, check them against a known phishing/malware domain list, and short-circuit if there's a hit. You save model calls and get a verdict the user can verify.

Disclosure: I work on WhoisFreaks. Our Threat Intelligence Feed is a daily list of phishing, malware and spam domains (plus C2/bot IPs) with confidence scores, if you want a data source for that check: whoisfreaks.com/products/threat-intelligence-feed