Scam calls are incredibly common, but Google has a new plan to detect them in real time using on-device AI models. It’s coming first to Google Pixel phones.

Google has announced the release of Scam Detection, a new real-time protection feature designed to shield users from scams and fraud attempts. The feature was previously announced during Google I/O 2024, but it is now making its way to users at long last. Scam calls are a big issue, with around $1 trillion in losses from these scams every year. The announcement was made as part of a wider slate of security features, along with the announcement of improvements to Google Play Protect.

How will it help, exactly? Scam Detection will make use of on-device AI to analyze conversation patterns and identify potential scam calls in real-time. If a call exhibits suspicious characteristics, such as a caller impersonating a bank representative, a carrier rep, or otherwise demanding urgent fund transfers, the feature will issue an audio, haptic, and visual warning to the user.

According to Google, Scam Detection has been engineered with user control in mind. The feature is deactivated by default and can be enabled or disabled at any time through the Phone app settings. If you do want to enable it but you’re worried about someone at Google potentially snooping on your convos, all AI processing occurs locally on your phone, ensuring that no conversation audio or transcripts are stored on the device or transmitted to Google servers.

The AI model powering Scam Detection is Gemini Nano, Google’s latest on-device AI innovation, which will be featured on Pixel 9 series devices. It should be noted that while Pixel 6 and above users don’t have Gemini Nano, they will also benefit from this feature, thanks to other robust on-device machine learning models. We don’t know yet if it will work identically or if there will be differences with the Gemini Nano-powered version.

The rollout of Scam Detection begins today for English-speaking Phone by Google public beta users in the U.S. with a Pixel 6 or newer device. As with every AI feature, it might take some time to actually become reliable, and we might see things such as false positives triggering a warning. Any issues should be ironed out within the next few months.

Source: Google