Indicates the type of a location. Location Type Defined by CCTVPLUS Indicates the role that a specific instance of a description takes among all descriptions. Description Role Defined by CCTVPLUS The role in the overall content stream. Part Role Indicates a kind of content rendition. Content Rendition Defined by CCTVPLUS CCTV Newscontent Co.,Ltd. CCTV Newscontent Co.,Ltd. All Rights Reserved. No access Chinese mainland CCTV Newscontent www.cctvplus.com +86-10-83949886 +86-10-63978447 contact@cctvplus.com 2026-08-12T09:50:34Z 2026-08-12T09:50:34Z 2026-08-11 2014-11-06T09:22:45Z China Central Television China Other Recent/File China-AI Models/Global Experts Open-source AI is irreversible global trend: experts Western and Chinese experts see open-source AI models gaining global traction as they now rival closed-source systems in technology, making open development an irreversible trend. The experts' views come as the U.S. government is reportedly moving to exempt open-weight models from safety reviews. In June, U.S. President Donald Trump signed an executive order aimed at managing security risks of AI models, allowing the government to obtain access to specific AI models up to 30 days before their release for review purposes. The U.S. federal government plans to subject only closed-source models to safety reviews, according to sources familiar with a top-level meeting at the White House on Aug 4, as relayed by Western outlets. Open-weight models will escape scrutiny, even when they are developed by Chinese companies. Oliver Buchmueller, a professor at Imperial College London, noted his recent use of China's Kimi and pointed out that open-source models provide global access. "Myself, I've been subscribing just very recently to Kimi. Many of the wit models are open-source models. It really is an access for the entire community and entire world," said Buchmueller. He is also a senior researcher at the European Organization for Nuclear Research (CERN). Cai Yunfeng, a researcher at the Beijing Institute of Mathematical Sciences and Applications (BIMSA), observed that open-source models are now technically on par with closed-source ones, with only a few months of performance gap. "Currently, there is almost no technological gap between open-source and closed-source models, though there is a performance gap of about three to six months," said Cai. Joe Neeman, researcher at the Machine Learning Lab, University of Texas at Austin, argued that even dominant closed models cannot stop the inevitable rise of open development. "Even if you have one very successful and very rich company keeping itself in a closed model, the world is too big. And I think eventually the open development is inevitable," said Neeman. Chinese/English/Part Mute FILE: China - Exact Location and Date Unknown (CCTV - No access Chinese mainland) 1. Various of servers 2. Computer keyboard, person typing 3. Animations of data 4. Animation of AI chip Beijing, China - Recent (CGTN - No access Chinese mainland) 5. SOUNDBITE (English) Oliver Buchmueller, professor, Imperial College London; senior researcher, European Organization for Nuclear Research (CERN) (partially overlaid with shot 6/ending with shot 7): "Myself, I've been subscribing just very recently to Kimi. Much of the wit models are open-source models. It really is an access for the entire community and entire world." ++SHOT OVERLAYING SOUNDBITE++ FILE: China - Exact Location and Date Unknown (CCTV - No access Chinese mainland) 6. Computer screen showing Moonshot AI's Kimi model ++SHOT OVERLAYING SOUNDBITE++ Beijing, China - Recent (CGTN - No access Chinese mainland) 7. Oliver Buchmueller talking with reporter FILE: China - Exact Location and Date Unknown (CCTV - No access Chinese mainland) 8. VArious of screens showing Moonshot AI's Kimi model, DeepSeek's V4 model 9. Various of ChatGPT webpages on cellphone screen Beijing, China - Recent (CGTN - No access Chinese mainland) 10. SOUNDBITE (Chinese) Cai Yunfeng, researcher, Beijing Institute of Mathematical Sciences and Applications (BIMSA): "Currently, there is almost no technological gap between open-source and closed-source models, though there is a performance gap of about three to six months." FILE: Guiyang City, Guizhou Province, southwest China - Date Unknown (CCTV - No access Chinese mainland) 11. Various of server racks Beijing, China - Recent (CGTN - No access Chinese mainland) 12. SOUNDBITE (Chinese) Joe Neeman, researcher, Machine Learning Lab University of Texas at Austin (ending with shot 13): "Even if you have one very successful and very rich company keeping themselves in a closed model, the world is too big. And I think eventually the open development is inevitable." 13. Joe Neeman talking with reporter FILE: China - Date Unknown (CCTV - No access Chinese mainland) 14. Various of intelligent bionic robots working