Naver said on Tuesday it has completed an upgrade to its malicious comment detection system, AI Cleanbot 3.0, to strengthen detection of life-disparaging and secondary-harm expressions.

The upgrade focuses on blocking comments that encourage disregard for life, including references to suicide, death and bodily harm. It also targets mocking, disparaging and hateful comments aimed at victims of incidents and accidents, as well as their bereaved families. The system was enhanced to detect malicious intent based on context by analysing not only news comments but also article titles and bodies.

Introduced in 2019, AI Cleanbot expanded its standards in 2020 with sentence context-based detection. It has continued to broaden its detection scope, including by reflecting the Korea Internet Self-Regulation Organization's (KISO) self-regulatory policy guidelines on hateful expressions in 2023, covering sexually offensive content, hateful, disparaging and discriminatory expressions, and evasive malicious comments.

Naver currently operates policies this year including not providing comments at the bottom of the main text in the politics and election section, and automatically disabling the comment service on an article when malicious comments exceed certain criteria.

Kim Soo-hyang (김수향), a Naver leader, said the company is continuously enhancing Cleanbot's performance to detect not only profanity and vulgar language but also newly emerging hateful, disparaging and discriminatory expressions. Kim said it will further improve performance by listening to a range of opinions, including by focusing on blocking content that encourages disregard for life and mocks or expresses hate toward victims and their bereaved families.