The real-time detection of Google phone scams is now available for Pixel devices.

One of the AI-driven features of Gemini that Google introduced at its I/O event this year is a solution for annoying scam calls. This tool has the capability to identify if a call is suspicious in real-time and can alert the user to hang up as soon as possible. Currently, this call fraud detection feature is available for Pixel 6 devices and later, as long as they are part of the Phone by Google public beta program in English.

The company mentioned that scam calls are evolving and becoming increasingly sophisticated, making them harder to detect. Fraud Detection uses on-device artificial intelligence to assess in real-time whether a call is potentially a scam. For example, if the caller presents themselves as a representative from your bank and asks you to transfer funds to another account under the pretext that yours has been compromised — a common scam tactic — you will receive an audible alert and a vibration. Looking at your phone, you will see a visual warning and a button that allows you to easily end the call. In case the AI makes a mistake, you can select the "Not a scam" button.

This Fraud Detection feature is turned off by default, and it is up to the user to decide whether to activate it. Google assures that it does not send your calls or their transcripts to a remote server, as call processing is done directly on the device. In the Pixel 9 series, this feature is powered by Gemini Nano, which Google describes as its most efficient model for on-device tasks. For older Pixel devices than the Pixel 9, other machine learning models from the company are used.

Although Google has not specified when real-time fraud detection will be available outside of the beta phase, it promised that it will soon be implemented on more Android devices. In October, the company also launched an improved version of scam detection for text messages, which uses on-device machine learning models to identify fraudulent texts.