The center now sends nearly 10,000 cross-border parcels a day to more than 30 countries and regions. As volumes rise, export efficiency keeps improving: the consolidation cycle has shrunk from two or three days to one, and queue times for logistics vehicles are down nearly half.

A nationally pioneering 'cross-border list plus declaration' and 'inspect first, load later' digital consolidated-container model lets cross-border e-commerce, general trade and market-procurement goods ship in the same container, so firms need not wait for同类 goods to accumulate - saving about 500 yuan per container. TIR road transport enables cross-jurisdiction 'arrive, inspect, release,' while land-air and sea-rail transshipment open diversified routes.

Smart customs upgrades - intelligent gates, electronic seals and round-the-clock zones - let parcels be precisely judged on passing the scanner, cutting manual steps. After declaration and inspection in Yiwu, shipments are sealed with secure smart locks with fully traceable data, and can be auto-transferred at Ningbo or Shanghai ports, removing the time limits of manual seal checks and lifting goods-flow efficiency.

Image: Tourist goods at Yiwu International Trade Mart. Source: Wikimedia Commons (CC BY-SA 4.0).

Discussions centered on three topics: science and technology innovation to help SMEs transform; digital and intelligent empowerment to raise SME digital literacy and help them ride the digital and AI wave into global value chains; and inclusive development so more SMEs seize financing, technology and market opportunities. Ministers stressed that AI can broaden participation in the formal economy and help SMEs respond faster to supply-chain shocks.

China proposed establishing an Asia-Pacific AI Empowerment Center for SME Development in Guangzhou and building an Asia-Pacific SME Cooperation Zone in Shenzhen. The meeting issued the Chair's Statement and the Guangzhou Initiative on SME Innovation and Resilience, contributing China's experience and innovations.

In closing remarks, Li Lecheng urged turning the day's discussions into practical action, calling for small, fast, light and precise digital products and AI applications to accelerate AI empowerment of SMEs. Experts noted AI is a 'general-purpose technology' that, for SMEs, is 'both a challenge and an opportunity' - letting small firms rent previously unaffordable 'brainpower' and, unburdened by legacy systems, rebuild as 'small organization, big capability.'

Image: Efficiency of AI-related computer chips. Source: Wikimedia Commons (CC BY-SA 4.0).

The agent's logic rests on two wheels. The 'AI capability plus real-owner ecosystem' helps users choose: rather than listing parameters, it weighs space, range, charging and family needs to give a leaning recommendation with stated trade-offs, drawing on Autohome's Cangjie industry model plus 1 billion real reviews and 50 million owner testimonials from its 150-million-owner base. The 'AI capability plus service ecosystem' helps users own: natural-language requests are matched to fulfillable resources.

On the supply side, Zhishi connects nearly 30,000 4S stores, more than 12,000 used-car dealers and over 100,000 upkeep-service resources across 369 cities, with qualification checks, capability verification and credit-rating admission to surface quality merchants. Its value proposition is distilled to two lines: less hesitation when choosing, less fuss when owning.

Autohome chairman and CEO Liu Chi framed the launch as a shift from information platform to trusted decision-making, while CTO Xiang Bibo said AI excels at processing information but information volume does not equal answer credibility. As L4 autonomy arrives, the agent is designed to extend from serving owners to serving the vehicle itself - arranging its own charging, washing and maintenance.

Image: Vehicles on display at an international mobility show. Source: Wikimedia Commons (CC BY 4.0).

The shift is driven by price-performance. UBS analysts note that as frontier-model costs climb, overseas users have grown ROI-conscious; China's models combine systematic price advantage with rapidly closing capability gaps. On Artificial Analysis, Kimi K3, Zhipu GLM-5.3, Alibaba Qwen 3.8 and DeepSeek V4 Pro sit just behind the top closed US models, yet Kimi K3 costs about $0.84 per task versus $3.69 for Claude Fable 5.1 - roughly a 4.4x gap.

Foreign firms are taking notice: US food-delivery giant DoorDash, Airbnb and Germany's Siemens have begun shifting back-office AI to Chinese models. Moonshot is negotiating revenue-share deals with Microsoft, Amazon and Google to host Kimi K3, seeking a cut as high as 30%, while Zhipu has started distribution cooperation with AWS and is talking with two large Western cloud providers.

To balance open release with monetization, vendors are tightening licences: Qwen 3.8-Max and Kimi K3 require separate permission for MaaS or AI-assistant operators above certain revenue thresholds. Per Xinhua, overseas AI firms including Harvey (built on Kimi K3) and Thinking Machines (using DeepSeek-V3 routes) now treat Chinese open-weight models as a base for their own products - turning China's models from exported products into exported technology.

Image: Efficiency of AI-related computer chips. Source: Wikimedia Commons (CC BY-SA 4.0).

The 70M-parameter model is pretrained entirely on large-scale synthetic data generated from a structural causal model (SCM), covering varied data scales, variable types and functional dependencies. Its architecture introduces dual-stream feature grouping, a lightweight attention residual and sparse mixture-of-experts to model feature relations at multiple granularities, while Test-Time Scaling boosts prediction by adding inference compute without changing pretrained weights.

On four public benchmarks - TALENT, TabArena, BCCO and OpenML-CTR23 - Xiaomi-TabLDM reached the first tier. It ranked first on OpenML-CTR23 regression and on TALENT binary classification, and second on regression in the other three. On TabArena regression it posted the second-highest Elo score while cutting training time 82% and prediction time 68% versus the top-ranked TabFM.

In real industrial checks it lifted material-property prediction accuracy by 130% with no fine-tuning and cut无效 trial molding by about 90%; versus XGBoost it lowered failed-sample ratio by about 31% in part-weight prediction and average error by 54% in production-composition prediction, needing only about 30 new samples to cut error a further 62% after a condition change. Weights and code are open on GitHub and Hugging Face with a scikit-learn-compatible interface.

Image: Efficiency of AI-related computer chips. Source: Wikimedia Commons (CC BY-SA 4.0).

Bank of America has been appointed overall coordinator, working with sponsors CICC, Deutsche Bank and Goldman Sachs. People familiar with the matter said the listing could happen as early as this year, though size and timing remain subject to change. The company was reported to have closed a $3.5 billion funding round at a $35 billion valuation in late July and is in talks for a new round at a roughly $50 billion pre-money valuation.

The filing follows the July 16 launch of Kimi K3, a 2.8-trillion-parameter open-weight model with a 1-million-token context window that quickly drew global attention; Moonshot's annual recurring revenue roughly tripled in three months to about $300 million. To list, the company unwound its offshore red-chip structure and redomiciled in mainland China, a pre-listing move now expected of Chinese AI firms seeking foreign listings.

Investors are watching whether institutional money still wants exposure to Chinese frontier models despite US export controls. A thin order book would signal caution; a strong one would mark a real vote of confidence in Chinese labs' ability to close the capability gap at a fraction of the price - API licensing already accounts for more than 70% of Moonshot's revenue.

Image: A humanoid robot representing the frontier of Chinese AI. Source: Wikimedia Commons (CC BY-SA 4.0).

The model carries 320 billion total parameters with 18 billion active per token, designed for low inference cost. Zhipu says GLM-5.3-Flash scores 57 on the Artificial Analysis Intelligence Index - level with Anthropic's Claude Opus 4.8 - while costing about one-tenth as much to run as GLM-5.2 and roughly one-fortieth of Opus 4.8.

Crucially, Zhipu disclosed that during a large stealth trial - when the model appeared anonymously as 'Ox Alpha' on OpenRouter and OpenCode - all request traffic was served by a cluster of more than 100,000 domestically produced chips, linked by a self-developed high-bandwidth interconnect, with no Nvidia hardware involved. The trial processed about 62 trillion tokens, and in its first three days on OpenRouter alone handled more than 11 trillion tokens.

The architecture combines sparse and linear attention, cutting attention computation by about 3x and context-cache memory by more than 4x versus GLM-5.3 while supporting a 1-million-token window. Zhipu says end-to-end service performance on the same hardware improved about 3x versus the initial baseline, with hardware efficiency and per-token cost approaching those of mainstream Nvidia GPUs - evidence that Chinese chips can economically support frontier-scale inference at global usage volumes.

Image: Efficiency of AI-related computer chips. Source: Wikimedia Commons (CC BY-SA 4.0).

The cost is showing up in the data. The State Administration for Market Regulation said 50 recall announcements covering 26 brands were issued in the first half of 2026, involving about 1.65 million vehicles, of which domestic-brand recalls reached 308,000 units - up 681.2% year on year. New-energy recalls rose in frequency, with defects concentrated in software logic and power-battery issues.

Regulators have responded with a combined punch. On August 26, at a State Council Information Office briefing on the 15th Five-Year Plan, MIIT vice minister Xin Guobin said radical 'innovation' that skips adequate testing before fitment must be curbed, and that products failing full validation must be kept off the market. In August, MIIT, with the ministries of public security, ecology and environment, and market regulation, launched a one-year special campaign on motor-vehicle production consistency and quality improvement.

The campaign builds a development-cycle oversight mechanism, imposes key scrutiny on models moving from kickoff to production in under 18 months, and can delay announcement filings where test materials are incomplete. Proposed revisions to national standards would raise reliability road-test mileage for EVs, plug-in hybrids and fuel-cell vehicles to no less than 30,000 km, aligning them with combustion-engine rules and raising the bar for 'rush-built' NEVs.

Image: Vehicles on display at an international mobility show. Source: Wikimedia Commons (CC BY 4.0).

Compared with the 9,210 stations reported on August 16, NIO added 112 sites in three weeks, 50 more swap stations and 365 more charging piles. On the same day, the company's 91st 'power tour' route - the Jinnan溯源 (origin-tracing) loop - was opened, spanning 1,849 km with 13 swap stations along the way, covering cities including Taiyuan, Jinzhong, Linfen, Yuncheng, Jincheng, Changzhi and Yangquan.

The fifth-generation swap station, first deployed in Quanzhou, Fujian at the beginning of August, has cut a single swap to about 1 minute 48 seconds and can serve up to 200 vehicles a day, now compatible with the NIO, Onvo and Firefly brands. NIO plans to expand its power-tour routes to 100 by year-end, with the Silk Road swap route to open in September.

Backed by 1.26 million cumulative deliveries, the network averages roughly 95 swaps per vehicle. Once mocked as too heavy and too slow to pay off, NIO's swap infrastructure is increasingly cited by rivals including CATL, BAIC and Geely as a direction worth following, turning a capital-intensive bet into a battery-asset management system competitors cannot quickly replicate.

Image: Next-generation EV charging infrastructure. Source: Wikimedia Commons (CC BY 3.0).

Hunan province stood out, with auto exports reaching 193,000 units in the first seven months, up 61.3% to a record high. Exports to ASEAN led the destinations, while shipments to Russia and Belarus grew more than 1.6-fold, illustrating a shift from simply 'selling products' to 'global operation.'

The trend underscores that auto export growth is broadening beyond the coastal leaders to inland manufacturing provinces, with diversified regional markets - ASEAN, Russia-Belarus and others - cushioning external volatility.

Image: Vehicles showcased at an international mobility show. Source: Wikimedia Commons (CC BY 4.0).

Under a core-technology list system, the province is tackling 'chokepoint' technologies such as automotive-grade chips and all-solid-state batteries, while deepening 'vehicle-component coordination' and cross-region 'fly-in' cooperation to strengthen the local supply chain.

The move reflects how major provincial economies are betting on NEVs as a pillar of advanced manufacturing, competing not just on assembly but on the harder layers of chips and next-generation batteries.

Image: Solar power installation representing clean, distributed energy. Source: Wikimedia Commons (CC BY-SA 4.0).

Sichuan has taken the lead by launching five regional recycling centers, and a management rule from six ministries including the Ministry of Industry and Information Technology took effect in April 2026. Yet more than 60% of retired batteries still flow into small, informal workshops, with lagging collection outlets and poor channels remaining the main pain points.

The gap highlights both a risk and an opportunity: as the first wave of NEVs reaches end-of-life, building a formal, traceable recycling chain is becoming a strategic segment of the battery industry, not just an afterthought of manufacturing.

Image: Low-speed electric vehicle charging scene in Jiangsu, China. Source: Wikimedia Commons (CC BY-SA 4.0).

Wang also clarified a market rumor, stating that CATARC has not judged the more than 320 enterprises laying out solid-state batteries and that the claim that 'only nine fully passed all compliance items' was false - pushing back against speculation that had unsettled the sector.

The standards work signals that as solid-state batteries move from lab to line, regulation is racing to keep pace, providing a clear compliance framework for a technology still in the engineering-validation stage.

Image: BYD 1,500 kW flash charging station demo, representing next-generation battery infrastructure. Source: Wikimedia Commons (CC BY 3.0).

Reduced effective capacity from dual risks in the Hormuz Strait and Red Sea has driven the spike. As a bellwether of small-commodity exports, Yiwu has seen visibly weaker orders, with the thin-margin small-goods sector hit hardest by the inversion of freight cost and cargo value.

In response, exporters are accelerating overseas-warehouse deployment and multimodal transport to buffer volatility, while some are rerouting or shifting product mix toward higher-value goods to absorb the higher logistics cost - a reminder that external shocks still test the resilience of the world's largest small-commodity hub.

Image: Overview of Yiwu International Trading Town, the world's largest small-commodity market. Source: Wikimedia Commons (CC BY-SA 4.0).

The guideline requires written informed consent from research participants and focuses on ethical issues in algorithm training and model evaluation - particularly the risks of hallucination and distortion in synthetic imaging data. It stresses that any AI-generated content touching health, body or property must be traceable, human-reviewed and disclosed to the user.

For enterprises, the message mirrors broader AI-governance principles: in scenarios involving health, finance or sensitive personal information, AI output should pass source-traceable, human-review and user-informed checks before it reaches the public.

Image: Efficiency improvement of AI-related computer chips. Source: Wikimedia Commons (CC0).

The standard requires that the path to a human agent be clear and not hidden in layers; simple questions go to AI while complex disputes flow to humans promptly; and context must sync across switches so consumers do not repeat themselves. Critically, enterprises may not use 'the AI answer does not represent the company' as an excuse - an AI reply on an official channel is the operator's expression of intent.

The China Consumers Association, which on August 31 named AI agents that gave inconsistent prices or evaded responsibility, noted that after-sales complaints made up 26.79% of all complaints in the first half of 2026, with 'false AI promises' and 'hard-to-reach human agents' most prominent. Consumers are advised to keep screenshots and failed-transfer records and may complain via 12315.

Image: Efficiency improvement of AI-related computer chips. Source: Wikimedia Commons (CC0).

As of the end of June, China had nearly 50 million electric vehicles on the road. Academician Ouyang Minggao has argued that vehicle-to-grid interaction is the key path to solving charging-capacity constraints, and the industry is entering a golden window for scaled commercial use.

The plan frames EVs not just as loads but as distributed, movable storage that can feed power back during peaks and charge during valleys, turning millions of car batteries into a flexible grid resource - a model already being piloted in several cities.

Image: Solar power installation representing clean, distributed energy. Source: Wikimedia Commons (CC BY-SA 4.0).

The report notes that the auto manufacturing profit margin fell to a historic low of 3.2% in the first quarter of 2026, with the price war's backlash clearly visible. Analysts argue that differentiation and overseas expansion are the main paths out of the squeeze, as homogeneous competition erodes margins across the board.

The assessment reframes the industry's contest from a sales-volume race to a profit-and-loss race, pressuring makers to prove sustainable unit economics rather than just scale - a theme echoed across this year's half-year earnings season.

Image: Electric bicycle and scooter charging station in Wuhan. Source: Wikimedia Commons (CC BY-SA 3.0).

Dongfeng announced that its oxide-polymer composite semi-solid battery, with an energy density of 350 Wh/kg, enters mass production in September, debuting on the eπ007 with a CLTC range beyond 1,000 km, 4C ultra-fast charging, over 72% capacity retention at minus 30°C, and a presale price of 150,000-200,000 yuan - bringing semi-solid tech into the mainstream family-car price band for the first time.

Other carmakers are following: GAC's 400 Wh/kg quasi-solid-state pack is planned for small-batch installation on its premium Trumpchi Haobo brand this year; SAIC MG has a 100,000-yuan-class semi-solid model; CATL's condensed (high-end semi-solid) battery at 360 Wh/kg delivers in the third quarter; and NIO's ET9 will get a 150 kWh semi-solid pack in the fourth quarter with over 1,200 km CLTC range. Full-solid-state, by contrast, still carries a production cost more than four times that of liquid batteries, with a 1 yuan per Wh scale threshold still ahead.

Image: BYD 1,500 kW flash charging station demo, representing next-generation battery infrastructure. Source: Wikimedia Commons (CC BY 3.0).

Analysts link the shift to two rounds of fuel-price hikes in July totaling 985 yuan per ton, which steadily tilted consumer willingness toward electrification. The reading underscores how decisively the mainstream Chinese market has swung to NEVs.

The figure also reflects a structural change: with NEVs now the default choice for most buyers, fuel vehicles are increasingly confined to niche and replacement demand, reshaping how automakers allocate production capacity and dealer networks.

Image: Vehicles showcased at an international mobility show. Source: Wikimedia Commons (CC BY 4.0).

The plan calls for expanding the supply of AI terminals, smart wearables, green home appliances, green building materials and new energy vehicles, and supporting consumption of green and smart goods - cultivating new consumption growth points by upgrading supply.

For the auto and AI sectors, the measure reinforces the policy direction of pairing intelligent, low-carbon products with demand-side incentives, supporting both China's NEV leadership and its fast-growing consumer-AI hardware race within a single coordinated framework.

Image: Efficiency improvement of AI-related computer chips. Source: Wikimedia Commons (CC0).

A surgical robot developed via HIT's advanced technology research institute has completed 15,000 abdominal surgeries with operation precision at the sub-millimeter level, controlled by a doctor at a console. The institute has incubated 42 enterprises in training and spawned 32 new enterprises from HIT, with one surgical-robot firm named to MIIT's list of next-generation AI surgical-robot technology challengers.

In a northern hub port, Yantai Port uses AI to monitor key equipment around the clock, cutting fault-diagnosis time by 70% and raising operational efficiency by 15% with zero safety accidents, shifting from manual inspection to intelligent early warning - a concrete example of AI driving industrial 'smart' transformation.

Image: Unitree G1 humanoid robot, an example of China's AI-enabled hardware. Source: Wikimedia Commons (CC0).

The plan sets clear milestones: by the end of 2026, the national service-provider resource pool should exceed 2,000 entities forming a tiered梯队; by the end of 2027, no fewer than 3,000, supporting a full-chain AI application service ecosystem. Provinces with national AI innovation application pilot zones should each host at least 100 pooled providers by end-2027.

Four core tasks include building a unified national resource pool, raising industry service supply, driving scaled application of standardized 'small, fast, light, precise' AI products, and strengthening support through tools such as 'compute vouchers' and front-line deployment engineers. Leveraging BRICS and China-ASEAN mechanisms, the action also promotes quality AI application projects going global.

Image: Efficiency improvement of AI-related computer chips. Source: Wikimedia Commons (CC0).

CBS noted that a Geely factory it visited runs at 99% automation under AI-guided robotic arms, while Chery ships about 70% of its products overseas and Geely's first-half 2026 exports reached 474,000 units, up 158% year on year.

Electrek reported on September 3 that in August, electric vehicles outsold fuel cars in Australia for the first time, with 27,078 pure EVs sold; among the top eight best-selling EVs, all but two Tesla models were made by Chinese brands. The milestone shows Chinese EVs leading not just at home but across fast-growing overseas markets.

Image: Vehicles showcased at an international mobility show. Source: Wikimedia Commons (CC BY 4.0).

The results, titled "Iono-Elastomer Interface Enables Low-Pressure, Highly Stable Solid-State Batteries," were published in the top international journal Advanced Energy Materials. The work is described as a milestone in practical core technology for sulfide solid-state batteries, solving a key engineering pain point of low-pressure operation and taking a critical step toward industrialization.

Analysts noted that while the achievement is an important research advance, moving from a lab result to actual vehicle installation still requires multiple rounds of engineering validation, a reminder that solid-state commercialization remains a marathon rather than a sprint.

Image: Low-speed electric vehicle charging scene in Jiangsu, China. Source: Wikimedia Commons (CC BY-SA 4.0).

The Dragon Armor battery is a power-battery system solution led by Xiaomi Auto. In R&D and production, Xiaomi defines the product, leads pack design and development, takes part in cell design, and exercises full-process quality control over manufacturing. The battery will be equipped on the Pengcheng series, whose new models are scheduled for release on September 7.

Separately on September 4, Sunwoda said its subsidiary Sunwoda Power plans an equity expansion bringing in Li Auto, which will invest 2.65 billion yuan for 1.4 billion yuan of new registered capital, a 8.79% stake, with both sides to collaborate more deeply on power-battery R&D, manufacturing and quality management.

Image: BYD 1,500 kW flash charging station demo, representing next-generation battery infrastructure. Source: Wikimedia Commons (CC BY 3.0).

Drafted by the competent administrative authority, the measures are designed so that institutional supply precisely matches the actual needs of the industry, truly "standing firm, working in practice and being genuinely useful," according to coverage of the rollout.

As a pioneer in local AI legislation, Chengdu's approach is watched as a template for other cities weighing how to give the fast-growing sector clearer rules on innovation, governance and application, complementing national-level AI policies and the broader 'AI+' strategy.

Image: Unitree G1 humanoid robot, an example of China's AI-enabled hardware. Source: Wikimedia Commons (CC0).

Unlike conventional smartphones that make users adapt to the device, an AI agent phone turns the handset itself into an intelligent assistant that proactively understands human needs and helps get tasks done. Industry observers note that a good agent product must be both capable and know "when to stop and let the user confirm."

The approval marks a step from generic generative AI chat toward on-device agents that can execute routines, reflecting China's fast-moving consumer-AI hardware race as models, chips and form factors converge in new devices.

Image: Efficiency improvement of AI-related computer chips. Source: Wikimedia Commons (CC0).

The plan's most notable signal is a clear tilt from "courting the giants" toward "supporting the small." For the first time in such industrial policy, it explicitly encourages local governments to inclusively support micro-entities such as "one-person companies" and "super individuals," recognizing that as large models lower the barriers to R&D, design and content production, the minimum viable scale of a venture is being rewritten.

A second keyword is inclusive computing power. The plan deepens a special action to empower SMEs with affordable computing, pooling and elastically allocating resources through China's national computing platform to give startups efficient, low-cost access - compared by analysts to the "electrification" of the AI era, where only cheap, ubiquitous compute can unleash a wave of entrepreneurship.

Image: Efficiency improvement of AI-related computer chips. Source: Wikimedia Commons (CC0).

According to the Yiwu Market Development Group, stationery export orders rose 63% month on month from late August, with creative functional stationery and licensed IP products accounting for more than 40% of the category and becoming the core growth engine. One A4 folder blending national-style illustration with magnetic storage drew 500,000 sets of intent orders from Indonesian and Mexican clients in a single day, while a fluorescent study set based on a hit animation was booked by a Saudi buyer for three quarters of capacity.

Environmentally friendly recycled-material stationery more than doubled year on year, reflecting a global green shift in campus consumption. Yiwu has added 17 smart flexible stationery lines since spring 2026, compressing delivery to 72 hours, and its stationery exports have grown double digits for 11 straight months, keeping it China's largest stationery export cluster.

Image: Overview of Yiwu International Trading Town, the world's largest small-commodity market. Source: Wikimedia Commons (CC BY-SA 4.0).

Exhibitors reported strong demand: one pearl merchant said orders from six processors sold out on the spot, while buyers from Thailand and neighboring countries crowded booths to negotiate, sample and place orders. Merchants credited the Zhejiang provincial commerce department and the Yiwu China Commodity City group for providing "nanny-style" support from booth filing and customs declaration to on-site translation.

Data underlines Yiwu's scale: the International Trade City gathers about 80,000 business entities across 26 categories and roughly 2.1 million product varieties, exporting to more than 230 countries and regions. The city's "brand going global" program has laid out 97 overseas projects in 38 countries and regions spanning overseas sub-markets, warehouses and showrooms, while Chinagoods uses AI product discovery and multi-language sites and Yiwu Pay handles cross-border settlement.

Image: Tourist goods on display at Yiwu International Trade Mart. Source: Wikimedia Commons (CC BY 2.0).

BYD led wholesale volume with 433,384 units, followed by Geely at 173,675 and Chery at 115,558. Analysts noted that NEV wholesale growth far outpaced the broader passenger-vehicle market, confirming the segment as the core engine stabilizing China's auto sector.

In a separate update, CPCA's Cui Dongshu said world新能源汽车 sales reached 13.52 million units in the first seven months of 2026, lifting the global NEV share to 24.1%, with battery electric at 16.8%, plug-in hybrid at 7.3% and conventional hybrid at 8.1% - the last becoming a notable growth bright spot.

Image: Low-speed electric vehicle charging scene in Jiangsu, China. Source: Wikimedia Commons (CC BY-SA 4.0).

In a video speech, CATL chairman Zeng Yiqun sharply criticized the industry's "speed-to-market" chaos: more than 600 new car models have launched in China so far this year, nearly three per day, with some compressing test cycles and causing batches of power-battery faults. He noted that the Ministry of Industry and Information Technology has launched a one-year special campaign on production consistency and quality improvement.

At the opening ceremony, eight innovative technologies were unveiled together, including ultra-fast charging, high-energy mixed solid-liquid batteries with an energy density of 350 Wh/kg now entering scaled production, a fourth-generation high-compaction lithium iron phosphate chemistry exceeding 200 Wh/kg, and long-life battery-swapping batteries.

Image: BYD 1,500 kW flash charging station demo, representing next-generation battery infrastructure. Source: Wikimedia Commons (CC BY 3.0).

Under the standard, when a vehicle is powered on, the one-pedal mode may only slow the car down by default and must not bring it to a complete stop. Drivers must manually enable the stronger mode, which automatically reverts to the default each time the vehicle restarts, and a persistent, clearly visible prompt must remain on while the mode is active.

Industry observers say the rule ends the era in which carmakers strongly pushed single-pedal mode as a factory default, forcing manufacturers to recalibrate their brake-blending and kinetic-energy recovery logic so that safety does not depend on a driver's habit.

Image: Vehicles showcased at an international mobility show. Source: Wikimedia Commons (CC BY 4.0).

The Saudi Arabian Public Investment Fund (PIF) subsidiary HUMAIN announced on September 3, 2026, the launch of HUMAIN M3, an advanced Arabic large language model developed in partnership with Chinese AI company MiniMax. HUMAIN M3 is built on MiniMax's open-source flagship model MiniMax M3 and features 428 billion parameters as a Mixture of Experts (MoE) architecture. The model was further pre-trained on over 1 trillion Arabic-native tokens, achieving the highest average score across seven public Arabic benchmark tests among all frontier models evaluated. The launch represents a landmark moment for Chinese AI open-source models gaining adoption in sovereign AI development programs, with industry analysts describing it as a significant milestone in the global AI ecosystem. HUMAIN M3 is designed to support government, enterprise, and consumer applications across the Arab world.

New York City education officials announced a comprehensive ban on AI tool usage for approximately 900,000 students in the nation's largest public school district, effective from the start of the next school week. The ban applies to all students through eighth grade in New York City's public schools, prohibiting the use of AI-powered tools including chatbots, writing assistants, and AI image generators during school hours and on school devices. The policy is designed to protect foundational learning skills during critical developmental years and prevent over-reliance on AI-generated content. NYC Education Officials stated that the ban will be reviewed annually and may be adjusted as AI literacy becomes an increasingly important skill. The policy has sparked debate among educators, with some arguing it is overly restrictive while others say it strikes the right balance.

British autonomous driving startup Wayve officially launched its AI-powered autonomous ride-hailing service in London on September 3, 2026. The service marks the first commercial deployment of Wayve's end-to-end deep learning-based autonomous driving system in a major global city. The company announced plans to expand the service to Tokyo, Japan, by the end of 2026, which will test the scalability of its technology in one of the world's most complex urban environments. Wayve's approach relies entirely on deep learning rather than traditional rule-based systems, using cameras and sensors to navigate dynamically in real-world traffic conditions. The London launch comes amid growing competition in the autonomous vehicle sector, with Wayve positioning itself as Europe's leading challenger to US-based Waymo and Cruise.

Meta officially released Muse Spark 1.3 on September 3, 2026, describing it as the most capable AI model the company has ever produced. Meta's Chief AI Officer stated that the release represents the largest single leap in model performance in the company's history. The model demonstrates strong capabilities in coding tasks and autonomous agent operations, closing the gap with leading models from OpenAI and Anthropic. Muse Spark 1.3 is designed for both consumer and enterprise applications, with availability through Meta's AI platform and API. The release contributed to a broader rally in US technology stocks on September 3, with Meta shares rising more than 3% in after-hours trading. Industry observers view Muse Spark as Meta's most serious competitive entry in the foundation model space to date.

Anthropic is reportedly close to finalizing a $15 billion revolving credit facility ahead of a planned initial public offering. According to sources familiar with the matter, Morgan Stanley is leading the facility, with Goldman Sachs, JPMorgan Chase, and Citigroup also participating as major lenders. These same four banks are serving as the primary underwriters for Anthropic's IPO. The credit arrangement would significantly strengthen Anthropic's balance sheet ahead of going public, providing financial flexibility as competition in the enterprise AI market intensifies. The reported facility size of $15 billion would be one of the largest pre-IPO credit lines in technology history. Anthropic has not officially commented on the reports.

NVIDIA announced on September 3, 2026, that it had agreed to acquire AI open-source platform Hugging Face for approximately $12.93 billion. This marks NVIDIA's largest acquisition to date and represents a major consolidation in the AI infrastructure space. Hugging Face currently hosts over 18 million developers and researchers, with more than 3 million models, 500,000 datasets, and 1 million AI applications on its platform. Under the deal terms, NVIDIA will pay around $11.9 billion to Hugging Face investors and provide up to $1 billion in equity incentive programs to retain employees. NVIDIA CEO Jensen Huang confirmed that the company will maintain Hugging Face's open-source philosophy and continue operating it as an open platform for the global AI community. The acquisition is subject to regulatory approval and is expected to close in early 2027.