
Enhanced work efficiency and security, including real-time advice, gaze correction, and deepfake detection
AI-generated summary
Zoom is announcing technical updates to its platform and new AI strategies through its annual event 'Zoomtopia'.
(San Francisco = Yonhap News) Correspondent Kwon Young-jeon = "I am speaking too fast right now, so please speak a little slower."
In the future, if you become nervous and speak quickly during a video conference, you will be able to hear real-time advice like this from an artificial intelligence (AI) agent.
Video conferencing platform 'Zoom' announced that it will introduce 'Live Meeting Agent' at the annual event 'Zoomtopia 2026' held on the 20th (local time).
The live meeting agent acts like a private panel that only the user can see, and provides real-time guidance if the user speaks too fast or deviates from the topic of conversation.
When a sudden question arises, we search internal data to find an answer.
In meetings with attendees speaking various languages, AI voice replication technology was combined to maintain the user's voice while transmitting interpreted voices in other languages.
Additionally, even if the user's gaze slightly deviates from the camera, AI naturally corrects the gaze as if looking directly at the camera within a 17-degree range.
In terms of security, a 'deepfake risk detection' solution has been added to respond to the risk of manipulation due to the advancement of AI technology.
This is a paid function that detects and warns of abnormal signs, such as whether the video or audio transmitted on the screen during a meeting has been manipulated or synthesized using deepfake technology.
We have strengthened the 'My Notes' function, which keeps track of meeting contents even if you leave the meeting midway or are unable to attend, to provide customized meeting minutes and summaries.
In addition, we decided to open a Model Context Rule (MCP) server that can be linked with Antropic's 'Claude' and OpenAI's 'ChatGPT' to support the ability to schedule meetings or summarize and search meeting minutes in an external AI environment without directly accessing Zoom.
Russell Dicker, Chief Product Officer (CPO) of Zoom, said, "The Gallup survey revealed that users do not want new AI tools, but rather to have their work completed smoothly," adding, "So we integrated AI into the system, including customers' daily meetings, conversations, and marketing events."
AI outlook — possibilities, not facts
Expansion of external AI model linkage function within the Zoom platform
Very likely · Within months

On the 20th, a power problem at the London Google Cloud data center paralyzed financial, medical, and postal services in the UK for six hours. This incident has once again raised concerns about cloud dependence concentrated on a few large companies.

About 40 major websites using the 'kp' domain, including North Korea's Ministry of Foreign Affairs and the Korean Central News Agency, suffered access failures for about eight hours. Experts are weighing the possibility of a Chinese communication line failure or DDoS attack.

Due to a network failure at Google Cloud's London data center, connection and payment disruptions occurred at major financial services and postal services in the UK, including Revolut, Lloyds, and HSBC.
![[AI Prism] The use of Chinese products has increased rapidly in the development of domestic AI models.](/api/img?u=https%3A%2F%2Fimg.yna.co.kr%2Fphoto%2Fcms%2F2023%2F02%2F17%2F61%2FPCM20230217000061990_P2.jpg&w=320&q=72&f=webp)
Among the AI model outputs released by Korean developers this year, 49.3% were based on Chinese products. In particular, while Alibaba's Qwen was used the most, there were no responses in the corporate survey, creating a contrast.

Deputy Secretary of Defense Emil Michael, co-chair of the U.S. White House's 'Superintelligence Force' (SIF), warned that if AI companies do not cooperate with the government, various administrative measures, including export controls, will be used to control their actions. He cited companies such as Antropic as an example and argued that strict checks are needed rather than voluntary measures, and that the spread of China's open source model and corporate espionage are also concerns.

The publishing industry is expressing concern as the mass purchase of AI learning books that began overseas has been detected in Korea as well. Kyobo Bookstore and Yes24 are rejecting or carefully monitoring suspicious purchase requests from overseas companies, and publishing organizations are urging the establishment and legislation of a compensation system through fact-finding surveys and appeals. However, it is pointed out that there are limits to practical response due to difficulties in verifying copyright and lack of systems.