Google explains how it uses AI for spam detection in Google Messages

Google Messages AI spam detection feature: Google's Android operating system offers several in-built protections in order to protect users from unauthorised app installations and messages containing suspicious links aimed at stealing users' data including their personal and financial information. One of the features that Google uses to protect Android users is called 'Spam Protection'. While this feature has been around for quite sometime, Google has now detailed exactly how it uses artificial intelligence (AI) to detect spams in the Google Messages app and protect users from it.

According to a report by Android police, Google has updated its document on spam detection. The updated document details that the company uses encrypted messages to train its spam detection model to detect and filter out spam messages.

For the unversed, Spam Protection feature in Google Messages app protects users from spam messages by detecting incoming messages in real-time and filtering them out in a separate folder. This feature is turned on by default.

How Google Messages uses AI to detect and filter out spam messages?

The Google Messages app uses on-device machine learning models to detect patterns in a users' messages and detect spam messages. It also checks URLs in the received messages for spams. It does so by uploading these URLs to Google to ensure that it does not contain a malicious link.

"When Spam protection is on, Google Messages filters spam by (i) using machine learning models that operate on your device to detect known spam patterns in your messages and (ii) checking if the message contains a URL and if yes, uploading that URL to Google to verify that the link is not malicious," Google wrote in its support document. In its support document, the company also said that it processes the contents of encrypted messages to detect spam and 'use those signals to train AI models to improve spam detection'.

Notably, messages shared in the Google Messages app with RCS turned on are always end-to-end encrypted.