
84.9% of short-form sanctions are abusive language and hate... Sanctions for 449,000 open chat accounts
According to Kakao's Talk Hello report for the first half of the year, 72.9% of general chat sanctions were for fraud and impersonation, 84.9% of short-form sanctions were for swearing and hate, and 449,000 open chat accounts were sanctioned.
AI-generated summary
Kakao released the Talk Hello report for the first half of the year, and included statistics on Talk Siren phishing prevention functions and sanctions for violations of operating policies.
In the first half of this year, the number of accounts blocked by users who encountered phishing risk warnings on KakaoTalk reached 176,000.
In general chat, fraud and impersonation, such as romance scams and side job scams, accounted for 73% of all sanctions, and in short form, malicious comments such as profanity and hate speech accounted for nearly 85% of the reasons for sanctions, showing clear differences in the patterns of user damage and harmful behavior by service area.
◇ KakaoTalk blocks 176,000 suspicious phishing accounts… General chat sanctions 73% fraud/impersonation
According to the information and communication technology (ICT) industry on the 11th, Kakao released its first half Talk Hello report containing these details.
According to the report, from January to June 2026, a total of 176,000 accounts were blocked from opponents using Talk Siren warning, a phishing prevention function of KakaoTalk, including Globe, Stranger, and Fake Signal. This is an increase of approximately 6% compared to the previous half year.
User reports decreased by 17% to approximately 35,000.
This appears to be because users responded by stopping and blocking conversations rather than following the reporting process after receiving a warning about suspected phishing.
By type, the number of blocked stranger signals that appear when contacting unfamiliar accounts in Korea increased by 62% compared to the previous half-year.
According to post-sanction statistics, the number of accounts restricted in the general chat area for violating operating policies was 138,735.
Although it decreased by 16.8% compared to the previous half, the rate of action completed within one hour after receiving the report reached 99.6%.
By reason of sanctions, fraud and impersonation were the most common, accounting for 101,175, or 72.9% of the total.
In addition to existing romance scams and side job scams, many variant methods such as live broadcasting, kidnapping threats, and no-show scams were confirmed.
On the other hand, sanctions against illegal spam, including gambling, such as online casinos and slot sites, increased by 36.0% from the previous half to 19,454.
◇ Sanctions on 449,000 open chat accounts… 85% of short-form sanctions are malicious comments
A total of 448,853 accounts were sanctioned in open chat.
Inside chat rooms (messages and bulletin boards), ‘swearing, hate speech and hatred’ (39.2%) and ‘obscenity and sexual acts’ (25.2%) were the main topics.
In particular, after clarifying the standards for sanctions for buying and selling game items and accounts, the number of measures related to 'illegal products and services' surged by 322.1% compared to the previous half-year.
In the open chat cover/profile area exposed to an unspecified number of people, the proportion of sanctions for obscene content (59.4%) was more than half.
In the posts and comments section, which includes Talk profiles and short forms (Tap Now), 134,918 accounts, an increase of 64.5% compared to the previous half, were restricted.
In particular, among the sanctioned accounts (approximately 70,000) in the short-form area, a whopping 84.9% contained profanity, hate speech, violence, and hatred.
Most of the comments were slang and personal attacks that occurred in the comment window at the bottom of the video.
The artificial intelligence (AI) proactive action rate for short-form harmful content recorded 94.94%.
In the Talk Profile area, sanctions for 'violation of investment advisory policies' by impersonating stock reading room operators, etc., increased by 188.5% compared to the previous half.
Kakao said, “We plan to continue to strengthen our response to evolving new scams and malicious harmful content through advance warning systems, advanced AI detection, and cooperation with related organizations.”
AI outlook — possibilities, not facts
Kakao will strengthen its response to new scams and malicious content by upgrading its advance warning system and AI detection.
Likely · Within months

The Korea Robot Industry Promotion Agency will hold the 21st International Robot Contest at EXCO in Daegu on the 13th and 14th. 2,271 people from 1,303 teams, ranging from elementary school students to companies, will participate, and 54 events will be held, including a humanoid robot competition based on a supermarket environment. A total of 136 awards, including 4 Presidential Awards, and a prize money of 30 million won will be awarded.
As NVIDIA CEO Jensen Huang admitted to an affair with Alice Huang, a woman suspected of being a Chinese spy, suspicions were raised that Alice Huang abused this relationship to indirectly leak NVIDIA AI chips worth at least $7.5 billion to a Chinese company. It was discovered that 240,000 high-performance semiconductors, including H100, H200, and A100, were smuggled into China or provided on lease.

Samsung Electronics is collaborating with AMD to advance the next-generation semiconductor manufacturing test infrastructure. AMD unveiled the 6th generation Epic Embedded 9006 series processor and introduced a configuration combined with Samsung DDR5 memory, PCIe 6th generation SSD, and CXL 3.2-based MD310 module. It provides up to 256 cores and up to 1.8 times improved throughput compared to the previous generation, and is used in industrial environments such as semiconductor automated test equipment.

The city of Seoul will reorganize and operate 'Project Seoul' on the 20th, which provides integrated information on the progress of public architecture projects in each stage of competition, design, construction, and completion. AI technology review solutions will be introduced on a pilot basis and a dedicated window will be established to expand information accessibility for citizens and architects.

The Ministry of Science and ICT held the '2025 Bio Future Forum' at the Westin Chosun Seoul on the 11th and discussed AI bio innovation strategies and K-Moonshot promotion plans under the theme of 'Sovereign Bio: The future of artificial intelligence (AI) bio innovation opened by K-Moonshot.' Kim Tae-hyung, CEO of BioNexus, gave the keynote lecture, and plans to expand strategic investment in AI new drug development, BCI, gene cell therapy, and synthetic biology through the 7 major seeds and K-Moonshot national projects were announced.

South Korea ranked 10th among 70 countries in the IMD global digital and AI competitiveness ranking, up eight spots from the previous year. The U.S. took first place, followed by Singapore second and China seventh. South Korea ranked third in technology infrastructure and first in R&D expenditure per GDP.