South Korea operator KT developed an AI-powered phishing detection system which flags suspicious voice calls in an attempt to combat the growing threat of phone scams.

KT’s head of network centre Lee Jong-sik noted it is strengthening AI technology research to prevent customers from being exposed to increasingly intelligent voice phishing, adding it will continue to cooperate with government agencies to “take the lead in eradicating voice phishing damage”.

In a statement, the operator explained its voice phishing detection agent converts voice calls into text in real time and uses a small language model to detect statements related to financial fraud.

It currently being beta tested by the operator’s in-house team and is scheduled for launch later this month, after being added to spam blocking app WhoWho. The app is made by VP Inc, with a user base of 6 million.

The agent uses on-device processing, eliminating the need to send analysis data over the network, for increased personal information protection.

KT obtained anonymised voice phishing script data from the National Institute of Scientific Investigation to optimise voice recognition technology and the small language model.

In addition, its voice phishing suspected number notification service can detect phone numbers suspected of scams using an AI model and notify users in real time.