LG Uplus Launches On-Device AI to Combat Deepfake Voice Scams

South Korea's LG Uplus is set to deploy a real-time AI voice authentication tool on June 30, designed to detect manipulated or synthetic voices within five seconds of call initiation. The system analyzes over 3,000 hours of voice data—equivalent to approximately 2 million calls—to identify frequency distortions and unnatural pronunciation patterns characteristic of deepfake audio.

Unlike cloud-based solutions, the technology operates entirely on-device, eliminating the need to transmit sensitive voice data to external servers. This architecture addresses growing privacy concerns while maintaining detection accuracy. The tool integrates with national law enforcement voice databases to flag potential phishing attempts involving impersonation or fabricated emergencies.

The launch comes as global telecom providers grapple with a 237% increase in AI-enabled voice scams reported by Interpol in 2023. LG Uplus positions the feature as part of its ixi-O AI agent suite, marking a strategic pivot toward security-focused telecom services. Industry analysts note this could set a new standard for call authentication protocols worldwide.