#AIFrontierModelGovernanceWatch Having been wanting to write this for some time, well before Fable/Mythos issue...complex subject, and of course, much more to be said here, particularly on China side.....
The U.S. gov't has ordered Anthropic to suspend foreign nationals’ access to its advanced AI models, raising questions about the right of the gov't to intervene in private tech companies.
Read more from Paul Triolo @pstAsiatech.
https://t.co/Z5JobzHs5k
China Tightens Control Over Strategic Minerals With Sweeping New Rules
The rules, which took effect Monday, cover everything from mining rights, mineral exploration and extraction to stockpiling, environmental restoration, regulatory oversight and legal liability. They serve as implementing regulations for the revised Mineral Resources Law, which took effect on July 1, 2025.
And then there are October 9, 2025 rules, still coming....
https://t.co/WqBeNs0QGh
SiliconFlow capitalizes on this squeeze by optimizing chip performance, inference frameworks, and resource scheduling to improve compute efficiency and lower costs. The startup claims its solutions can increase language model speeds tenfold while cutting costs by 46%.
China’s top securities regulator pledged to welcome more domestic listings from AI developers as well as Hong Kong traded firms to bolster onshore capital markets https://t.co/d07ZK8NRev
On the sidelines of the Shanghai conference, He reflected on the years since the restrictions were imposed. When under pressure, she said, she often visits Dujiangyan, a 2,000-year-old irrigation system in Sichuan built by engineers who split mountains and reshaped rivers using the limited tools available to them.
Huawei’s big comeback tests limits of US chip controls
...shortage one reason Huawei’s ambitions abroad remain limited for now. But firm has already begun marketing systems that combine 950DT chips with DeepSeek’s V4 model to customers in Middle East and Central Asia.
https://t.co/gAOtbvP4Gl
Zhipu GLM5.2 narrows the performance gaps, ranking right behind Fable 5 & Opus 4.8 - and its open source/weight, and it costs a tiny fraction of the other two.
In the world of Darios and Sam’s, I’m grateful for the open source/weight community worldwide to keep the pressure on.
China Moves to Boost the Use of Yuan Globally
China will launch pilot program for offshore yuan foreign exchange trading
https://t.co/pIbXsZ6wDn via @WSJ
‘China’s Instagram’ Readies Hong Kong IPO That Could See It Valued at Over $70 Billion -
Known as Rednote, the company’s net profit could surpass $3 billion this year
https://t.co/Wk71PS9xoB
Semafor: Senior Trump administration officials had weighed how to structure potential government equity stakes in major AI companies before the government’s export controls on Anthropic further roiled the industry.... Treasury Secretary Scott Bessent favored using equity in AI firms to seed Trump Accounts, while Commerce Secretary Howard Lutnick’s preference was that any equity be directed to a type of sovereign wealth fund.
China’s Alipay gets AI-agent overhaul to challenge rivals Tencent and ByteDance
Fintech giant is seen leveraging its billion-user base to launch a conversational interface that coordinates 10,000 services, with safeguards
https://t.co/0MqTYvRZ5B via @scmpnews
The team trained the model using Chinese chip designer Moore Threads Technology’s MTT S5000 graphics processing unit and its Musa platform, an alternative to Nvidia’s CUDA, the company said in a blog post.
Can Chinese silicon replace Nvidia? Here are 5 AI models trained on local chips
A growing number of Chinese AI labs are experimenting with shifting earlier model training phases onto domestic chips
https://t.co/iGfnnYTxIn via @scmpnews
Earlier this month, a team at Peking University released EvoPhys-World, a 5D world model that simulates movements in physical spaces. The model took the top spot on Stanford University’s WorldScore benchmark.
In June, a research team from Huawei and the Shenzhen Loop Area Institute said it used Huawei’s Ascend 910C chips to conduct post-training for DeepSeek-V4-Pro.
In April, on-demand services giant Meituan invited users to test its new trillion-parameter AI model, LongCat-2.0-Preview. The company said both training and inference were completed entirely on a “domestic computing cluster”.
In January, Beijing-based Zhipu AI open-sourced its new image generation model, GLM-Image, developed alongside Huawei Technologies. The model was trained using Huawei’s Ascend Atlas 800T A2 server – powered by the Ascend 910 AI accelerator – and its MindSpore deep learning framework, according to Zhipu.
China’s AI outpaces global rivals but trails in trust, survey shows
Many people believe Chinese AI models are leading global tech race – even in countries considered key US allies – according to new poll by London-based consultancy.
Wait, who is winning here?
https://t.co/GbW1k5rwm7