Flash News

TREE NEWS update: White House economic adviser Kevin Hassett said he was glad to see OpenAI and Elon Musk support Anthropic CEO Dario Amodei’s post on AI, calling it a private-sector roadmap for AI. Hassett added that AI safety is a “solvable problem.”

TREE NEWS update: Asked whether the AI industry should slow down, US President Donald Trump said whoever wins artificial intelligence wins the future. Trump added that safeguards can be put in place, but that some voices are being overly negative. He did not announce any specific policy or measure.

TREE NEWS reports: TRM analyzed roughly $52.7 million across 198.9 million settlements on the x402 payment protocol and concluded that most of the activity is not coming from AI agents. The finding challenges the narrative that autonomous agents are already spending money online at scale. The research covers the protocol’s settlement data as of the study period.

TREE NEWS update: US House Speaker Mike Johnson said AI regulation should not be pushed through hastily, and that lawmakers should meet with AI leaders before any legislation is drafted. He also said a proposed $5,000 check plan floated by Donald Trump would not add to the debt.

TREE NEWS update: Anthropic’s chief executive is urging a slowdown in AI development to a safer pace, citing safety concerns. OpenAI CEO Sam Altman said he agrees with those safety concerns and confirmed the company will not hold an initial public offering this year.

TREE NEWS update: Zhipu announced on September 13 that it has completed about $5 billion in financing, comprising roughly $2 billion in share placement and about $3 billion in convertible bond issuance. Proceeds will be used for next-generation model research and development, computing infrastructure, and business expansion.

TREE NEWS update: Chinese Academy of Engineering academician Wu Jiangxing said at the Pujiang Innovation Forum’s space computing industry session on September 13 that very low Earth orbit (VLEO) is the last strategic new resource layer in Earth’s space. He said 2026-2030 is a critical five-year window for China to change lanes and must achieve a full leap from technical trials to large-scale application. Missing it would let others seize orbital, spectrum and international standard resources first.

TREE NEWS update: Shanghai officially launched the “Mingzhu Constellation Plan” on September 13 at the Space Computing Industry Ecosystem Forum of the 2026 Pujiang Innovation Forum. The plan’s long-term goal is to build a gigawatt-class space-based computing constellation, with in-orbit verification of domestically produced computing payloads, optical computing payloads and new energy systems, followed by initial satellite deployment and large-scale networking, aiming to complete a globally covered, space-ground coordinated, independently controllable space intelligent computing infrastructure and service network by 2030.

TREE NEWS reports: Oracle’s earnings disclosed that all renewed GPU compute capacity was resold at prices 20% above prior contracts, with most of the equipment already in use for over four years. The analyst said this rebuts Michael Burry’s argument that GPUs depreciate rapidly and is bullish for Nebius, IREN, Nvidia and the Neocloud compute sector.

TREE NEWS update: Meta Platforms chief artificial intelligence officer Alexandr Wang said on September 13 that the industry is at a critical moment as AI capabilities grow rapidly, and that alignment work must advance quickly to keep pace. Wang said “AI for everyone” means people can trust powerful AI to reliably pursue their goals without unwanted side effects, and called for the industry to act prudently and thoughtfully during this period.

TREE NEWS update: Researchers at the University of Manchester published a paper in the British Journal of Cancer saying that three-dimensional tumor models built in the lab with biomaterials, which mimic the real tumor environment in the human body, can help accelerate anticancer drug development. The researchers also said the technology still cannot replicate the body’s full complexity, so animal testing will remain important for disease research and for assessing drug efficacy and safety for the foreseeable future.

TREE NEWS reports: SK Group Chairman Chey Tae-won said the AI data center the group is building in Ulsan has expanded to nearly 900 megawatts, and that SK plans to announce more partnerships with global technology companies as the project advances. The Ulsan facility is being developed jointly by SK Group and Amazon AWS, with an expected investment of 7 trillion won ($5.2 billion) and operations targeted for the second half of 2027.

TREE NEWS update: Hyundai Motor Group said it is developing its own autonomous driving technology built around Atria AI, an end-to-end artificial intelligence model designed to perceive surroundings and make driving decisions. Mass production of Level 2++ vehicles equipped with Atria AI is targeted for the second half of 2029. The group plans to keep training the model on real-world road data and extend it across production vehicles, with a long-term goal of reaching Level 4 autonomy.

TREE NEWS update: China’s National Data Administration Director Liu Liehong chaired a symposium on embodied intelligence in Beijing on September 10, the agency said. Representatives from the Chinese Academy of Sciences’ Institute of Automation, the Beijing Institute for General Artificial Intelligence, JD.com, Lightwheel, and the Beijing Humanoid Robot Innovation Center discussed how data can support embodied AI development. The industry is increasingly data-driven, with demand for high-quality, large-scale data continuing to grow.

TREE NEWS update: China Telecom Group general manager Liu Guiqing called for deeper coordination between computing and power industries in a keynote at the 2026 China Computing Conference, held September 11-13 in Langfang, Hebei. He urged joint efforts on green power direct supply and source-grid-load-storage models to enable efficient supply of diversified clean energy.

TREE NEWS update: China’s national computing platform has achieved unified nationwide monitoring of computing resources, the Ministry of Industry and Information Technology said, with a “one network, one chessboard, one integrated” development framework for the country’s compute capacity now basically in place. The national-level platform has gathered more than 10,000 registered enterprise users, listed over 2,000 computing products, connected more than 300 large models, and accumulated billions of records of monitoring data.

TREE NEWS update: BRICS leaders endorsed the work of seven special working groups under the bloc’s industrial track and praised the smart manufacturing and robotics group’s efforts on smart factory development, solution sharing, standards cooperation and talent training, in the New Delhi Declaration from their 18th summit. The declaration recognized AI’s potential to drive growth and sustainable development, calling its use for economic growth and social welfare vital to a fair and prosperous future, especially for Global South countries, and affirmed the World AI Conference in Shanghai.

TREE NEWS update: OpenAI CEO Sam Altman said the company will not hold an IPO this year, calling the timing inappropriate and saying a listing would come no earlier than next year. Altman said OpenAI is focused on addressing safety concerns around AI, adding that even a 10% risk of the technology killing everyone before the decade’s end would be unacceptable.

TREE NEWS update: OpenAI CEO Sam Altman said he agrees with Anthropic CEO Dario Amodei’s view that the pace of frontier model development needs to be controlled. Altman made the remark in a statement, aligning himself with his rival lab chief on the need for guardrails on cutting-edge AI. Neither executive offered specifics on what controls they favor.

TREE NEWS reports: A week after launch, users of OpenAI’s GPT-6 Astra are complaining the model has been nerfed, saying it has gotten dumber. OpenAI’s previous model went through the same cycle of launch praise and degradation complaints in July. No response from OpenAI was reported.