2.0.0.0
majorVersion 2.0.0.0 adds AI Deep Scan, an AI phishing detection model that runs entirely in your browser and gives a second opinion on suspicious email.
What changed
- AI Deep Scan, a second opinion on emails that look suspicious but are not caught by the automated rules
- A DistilBERT model trained for phishing detection, about 67 MB, downloaded once
- Runs fully offline in your browser using WebAssembly
- Returns its own verdict, a confidence score and written reasoning
- No email content is sent to any server for analysis
The heuristic checks catch common phishing by looking for patterns they already know. The AI Deep Scan is for everything else - the email that looks a little off but breaks none of the rules.
This release adds that second opinion. The model is a DistilBERT network trained for phishing detection. It downloads once, at about 67 MB, and then runs entirely offline in your browser through WebAssembly. It returns its own verdict, a confidence score, and a short explanation of what it saw.
Like every other check in the extension, it runs on your machine. Your email is never uploaded to be analysed.