Grammarly: AI Writing Assistant and Grammar Checker App

AI Security Review

Evidence context: threat category none; evidence quality strong.

This extension is the official Grammarly writing assistant from a verified developer ([email protected]) with 39 million users. The security findings detected are entirely consistent with legitimate extension behavior and known false-positive patterns.

The single manifest finding flags the 'tabs' permission in manifest.json. This permission is necessary and expected for a grammar checker that needs to analyze text across web pages. Grammarly must read page content to provide real-time writing suggestions, making this permission functionally required rather than suspicious.

All six network findings originate from bundled vendor code in webpack chunks (src/js/8581.vendors.chunk.js, src/js/2887.vendors.chunk.js, src/js/1044.vendors.chunk.js, src/js/1974.vendors.chunk.js) and WebSocket transport files (src/inkwell/assets/CAPIWebSocketTransport-DztTrDg1-8WOYftOV.js, src/inkwell/assets/CAPIWebSocketTransport-DztTrDg1-DSQiisKE.js). These files contain third-party dependencies and communication infrastructure. The findings detect standard network operations—fetch calls, jQuery AJAX requests, and WebSocket connections—which are how any modern web application communicates with backend services. No specific suspicious domains were extracted from these network calls.

Critically, the analysis found zero malware signatures, zero obfuscation indicators, zero code-smell findings, and zero extracted indicators of compromise (IoCs). The absence of these high-confidence threat signals is significant. Real malicious extensions typically exhibit at least one of: malware signatures in code, obfuscation techniques, suspicious domain connections, or credential theft patterns. None of these are present here.

Strongest Counterargument: A skeptic might argue that 39 million users doesn't guarantee safety—malicious extensions can gain large followings before detection, and the 'tabs' permission could theoretically enable data exfiltration. However, this counterargument fails because: (1) Grammarly is a publicly traded company with a 15+ year reputation, making supply-chain compromise extremely unlikely compared to anonymous publishers; (2) the 'tabs' permission alone doesn't indicate malice—password managers, ad blockers, and productivity tools all legitimately require it; (3) there is zero evidence of suspicious behavior in the actual code—no hidden domains, no obfuscation, no credential handling outside expected patterns. Without specific malicious indicators, the permission is benign in this context.

The findings represent normal operation of a complex web application with bundled dependencies. This extension should be classified as a false positive.

Key Reasons

• Verified publisher (Grammarly) with 39M users and legitimate developer email
• Zero malware signatures, obfuscation, or suspicious IoCs detected
• Network findings in bundled vendor code are expected behavior
• 'tabs' permission is functionally necessary for grammar checking

False Positive Considerations

• Network findings in webpack vendor chunk files (bundled dependencies)
• Sensitive 'tabs' permission required for legitimate functionality
• High finding count from bundled code, not malicious behavior

Reviewed 2026-06-10; recommended action: suppress false positive; model confidence 95%.