Is Google's New AI Listening To Your Phone Calls?

Google would like you to believe that it only wants to keep you safe by offering a new feature that listens to your phone calls. At Shoreline Amphitheater today in Mountain View, California, Google representatives took the stage to announce a slew of new AI features. There's new Google Search AI, new versions of Gemini, and even an AI-powered Google Photos search. But one of the few AI features that seems like it will directly impact the user experience of Android is a Gemini feature that will listen to the contents of your phone calls and deliver an alert if the call is suspected of being a scam. For example, an alert would appear if the caller asked for your bank username and password, something a real representative from your bank would never do.

In theory, this fraud alert feature is meant to help people who might not be savvy enough to avoid these types of scams. But of course, it's not possible to deliver those alerts without real-time processing of what's being said on your phone calls. In a blog post about the feature, Google claims everything will happen on-device, with the call remaining private to you. The feature is also planned to be opt-in, meaning users will have to turn it on in order for it to start working. 

However, it's easy to be skeptical of Google in this case, and it's not easy to take their word on anything privacy-related. After all, Google's entire game is collecting your information to target ads back at you, and your phone calls are a treasure trove of personal information that Google would very much benefit from having access to. In other words, they might say everything is private, but they also have an incentive to capture as much user data as possible. So, is this feature actually secure?

The scam alert feature uses Google Gemini Nano, an AI model made to run on-device, so there may be no reason for alarm. After all, Google has shown that it can run speech-to-text processing on-device for Google voice search and Gboard without any latency, so using AI to analyze that speech for language commonly used by scammers is certainly something that could be accomplished with modern, AI-capable chips like those found in newer smartphones. But the fact that so many people's first instinct has been distrustful isn't rooted in pure paranoia.

At the time of writing, Google is settling a massive class action lawsuit after it was revealed that the popular Incognito Mode in Chrome wasn't actually all that incognito. Google has agreed to delete the information it collected from Incognito users. However, given that it allegedly misled consumers about the privacy of Incognito in the first place, maintaining some level of suspicion is logical. For obvious reasons, the type of people who need this feature are not the type of people who get into the nitty gritty of data privacy. So, for the more privacy-conscious, it may be worth keeping this feature turned off should it ever make its way into a phone you own.