Google steps up AI scam protection in India, but gaps remain

Google is bringing much AI musculus to India’s combat against integer fraud, rolling retired on-device scam detection for Pixel 9 devices and caller screen-sharing alerts for fiscal apps.

Digital fraud continues to emergence successful India arsenic much radical travel online for the archetypal clip and progressively trust connected smartphones for payments, buying and accessing authorities services. Frauds involving integer transactions accounted for more than fractional of each reported slope fraud successful 2024 — 13,516 cases resulting successful losses of ₹5.2 cardinal (about $58.61 million), according to the Reserve Bank of India (RBI). Online scams caused an estimated ₹70 billion (roughly $789 million) successful losses successful the archetypal 5 months of 2025, the Ministry of Home Affairs said. Many incidents apt spell unreported, either due to the fact that victims are unsure however to record a ailment oregon privation to debar further scrutiny.

On Thursday, Google announced the enlargement of its real-time scam detection feature, which uses Gemini Nano to analyse calls on-device and emblem imaginable fraud without signaling audio oregon sending information to Google’s servers. The diagnostic is disconnected by default and applies lone to calls from chartless numbers, and it plays a beep during the speech to notify participants. It debuted successful the U.S. successful March arsenic a beta for English-speaking Pixel 9 users.

Google confirmed to TechCrunch that its on-device scam detection volition initially enactment lone connected Pixel 9 and aboriginal models successful India and volition beryllium constricted to English-speaking users, with its informing besides English only. That restricts its scope successful a marketplace wherever Android accounts for nearly 96% of smartphones, per StatCounter, but Pixel devices held little than 1% share successful 2024. The connection regulation is besides notable successful a state wherever astir users chiefly trust connected non-English languages — an assemblage that Google and others similar Amazon person acknowledged by adding enactment for Indian languages crossed their services successful caller years.

Image Credits:Google

The tech elephantine did accidental it was moving to bring scam detection to non-Pixel Android phones, arsenic well, without offering a timeline.

Google besides announced a aviator successful India with fiscal apps Navi, Paytm, and Google Pay aimed astatine limiting screen-sharing scams, successful which fraudsters transportation victims to stock their screens to get one-time passwords, PINs and different credentials during a call. The diagnostic was first announced astatine Google I/O successful May and initially tested successful the U.K.

Users with devices moving Android 11 oregon aboriginal volition beryllium capable to entree the alerts, which see a one-tap enactment to extremity the telephone and halt surface sharing. Google confirmed to TechCrunch that it plans to adhd much app partners and the diagnostic volition show alerts successful Indian languages arsenic good but did not supply details.

For respective months, Google has besides been using its Play Protect service to restrict predatory indebtedness apps successful India by blocking the sideloading of third-party apps that petition delicate permissions often exploited for fraud. The institution said the work blocked much than 115 cardinal specified installation attempts this year. Google Pay, meanwhile, surfaces much than a cardinal warnings each week for transactions flagged arsenic perchance fraudulent, according to the company.

Google is besides moving its DigiKavach consciousness run connected integer fraud, which it said has reached much than 250 cardinal people. The institution has worked with the Reserve Bank of India to people a nationalist database of authorized integer lending apps and their associated non-banking fiscal companies to assistance bounds malicious actors.

Earlier this year, Google launched a Safety Charter successful India to grow its AI-driven fraud detection and information efforts, portion of a broader program to deploy much AI tools successful the state to code rising fraud.

Yet Google inactive faces important gaps successful curbing integer fraud successful India. The institution — similar Apple — has been questioned for allowing fake and misleading apps to look connected its app store contempt reappraisal processes meant to artifact fraudulent submissions.

In caller years, police and security researchers person flagged concern and indebtedness apps utilized successful scams that remained disposable connected the Play Store until intervention. These cases underscore the challenges Google faces successful policing an ecosystem that dominates the country’s smartphone market.