Naver upgrades Cleanbot to block victim-targeting comments

Naver applied an upgrade to its AI-based moderation tool, deploying AI Cleanbot 3.0 starting on April 29, according to reporting by ChosunBiz and finance.biggo. The upgrade adds a context-aware evaluation that examines the article headline and body together with the comment text to judge malicious intent, the coverage says. Both ChosunBiz and finance.biggo report the new release specifically targets expressions that trivialize or devalue life and "secondary-harm" comments that mock or belittle victims and bereaved families of incidents and accidents. ChosunBiz's AMP article quotes Naver leader Kim Su-hyang saying, "We are continuously strengthening Cleanbot's performance to respond to newly emerging expressions of hate, belittlement, and discrimination."

Reporting states the principal technical change is cross-document context analysis: instead of classifying a comment in isolation, AI Cleanbot 3.0 combines signals from the comment and the associated article (headline and body) to infer malicious intent, per ChosunBiz and finance.biggo. The sources also trace Cleanbot's evolution: Naver introduced the system in 2019, moved beyond keyword filtering after 2020, and incrementally added detection for insulting expressions without explicit profanity, sexualized harassment, hate and discriminatory content, and obfuscated text that uses symbols to evade detection. Finance.biggo notes Naver cited external guideline input from the Korea Internet Self-governance Organization (KISO) as part of its approach.

Editorial analysis - technical context: Platforms that add article-level context to comment moderation typically improve recall for context-dependent abuse but face new evaluation and labeling challenges. Context-aware classifiers require training data that links comments to article text, which raises annotation complexity and can increase annotation cost. They also raise the risk of higher false-positive rates when benign commentary adopts sensitive language in explanatory contexts. Practitioners implementing similar systems often combine rule-based heuristics, context encoders, and post-classification human review to balance precision and safety.

Editorial analysis: The upgrade illustrates a broader industry trend toward multi-turn or cross-document content moderation where the surrounding content is necessary to infer intent. For platforms with large news comment volumes, adding article context can reduce exposure to comments that cause secondary harm in high-risk coverage (for example, incidents involving suicide or severe injury). Regional moderators and civil-society governance groups in South Korea have been active on comment policy, and reporting that Naver consults KISO fits a pattern of public-private coordination on safety rules.

For practitioners and observers: watch for any Naver transparency reporting that quantifies the change - for example, metrics on comments removed, appeals, or measured false-positive rates - since those will indicate operational trade-offs. Also monitor whether context-aware filtering is extended beyond news comments to other UGC surfaces, how threshold rules (such as when comments are auto-disabled) are set, and whether KISO or regulators update guidance that affects automated moderation criteria. Industry observers may also follow any published technical notes or partnerships that reveal annotation methodology and model architectures for context integration.

ChosunBiz and its AMP copy report that Naver applies policy actions in parallel with the model upgrade: when malicious comments on an article exceed a threshold, the platform automatically disables the comment service, displays a guidance message, and shows a "Green Internet" campaign banner; coverage also says Naver has tightened comment operation on political and election-related articles. These operational rules are reported as platform measures in the same sources that describe the AI Cleanbot 3.0 upgrade.

Summary of sources used in reporting: ChosunBiz and finance.biggo provide the primary descriptions of the upgrade, its focus on life-devaluing and secondary-harm comments, historical development since 2019, and quoted remarks from a Naver leader; Korea Times published a summary notice of the same upgrade.