AI-Dguker Android: AI to Catch Scammers in Calls, Chats

Google has expanded its introduction of new tools for detecting artificial intelligence to protect Android users from more sophisticated telephone fraud and attacks on SMS. The company notes that these measures were taken in response to the campaign using AI, which has already caused more than 1 trillion dollars in damage according to the Global Anti-Scam Alliance data.

The new tools are particularly focused on combating “spoken fraud”, which start off innocuously but gradually lead victims into dangerous situations. Users are increasingly facing text attacks, where harmless messages are used to manipulate victims into giving away money or personal information. Scammers are also resorting to impersonating trusted organizations and using room substitution techniques.

Previously, Android’s protection against spam mainly involved blocking threats before communication began. However, scammers adapted their tactics to become harmful during conversations. In response, Google implemented AI models to monitor real-time behavior patterns suspicious and notify users of potential risks.

The new SCAM Detection feature in Google Messages enhances existing spam protection by analyzing SMS, MMS, and RCS messages for suspicious patterns. If a message is deemed potentially dangerous, the user will receive a warning with the option to block the sender or report them. The function is automatically enabled for conversations with unfamiliar numbers and processes data locally without sending messages to Google servers.

Google has already rolled out this protection in the USA, Great Britain, and Canada in English, with plans to expand to other regions in the future.

Another key innovation is protecting against fraud in calls, where real-time conversations are analyzed for signs of fraud. For instance, if a caller requests money through gift cards, the system alerts the user with a sound, vibration, and visual warning. Machine learning models tailored for different generations of devices were used in beta testing, with the Gemini Nano model performing the best and being extended to all Pixel 9+ devices.

/Reports, release notes, official announcements.