via The Verge
Meta Introduces On-Device AI Screening to Combat WhatsApp Scams
Meta has begun rolling out an AI-powered screening system designed to detect scams on WhatsApp, marking a significant step in the company's ongoing efforts to protect its billions of users from fraud. The new feature leverages an on-device AI model that runs directly on users' smartphones, aiming to identify potential scam messages in real time.
Unlike cloud-based solutions that process data on remote servers, this on-device approach offers several key advantages: it enhances user privacy by keeping message data on the phone, reduces latency, and allows the model to operate even without an internet connection. The model is trained to recognize patterns commonly associated with scams, such as urgent requests for money, suspicious links, or messages posing as trusted entities.
As of 2026, the feature is being gradually deployed to users globally, with Meta emphasizing that the AI operates locally and does not compromise end-to-end encryption. This aligns with the broader industry trend toward edge AI, where processing happens closer to the user to improve both security and user experience.
While the system cannot catch every scam, it serves as an additional layer of defense, complementing existing tools like user reporting and message filters. Meta has also updated its safety guidelines, encouraging users to remain vigilant and report suspicious contacts. Early feedback suggests that the AI screening has already helped reduce scam incidents, although Meta remains cautious about over-reliance on automated systems.
This initiative reflects a broader move within major tech companies to integrate machine learning into core product safety features, addressing the evolving tactics of cybercriminals. For WhatsApp users, the update promises a safer messaging environment, with the promise of future iterations that could further refine scam detection accuracy.