AI Industry Development & Policy Dynamics

According to Nvidia, D-Matrix announced on September 10 it will integrate Nvidia NVLink Fusion to connect its next-generation Raptor XPUs to Nvidia's AI infrastructure platform. The AI inference chipmaker will leverage Nvidia NVLink scale-up, Spectrum-X scale-out networking, and the MGX rack

According to the Foreign Ministry, spokeswoman Mao Ning rejected allegations made in a U.S. artificial intelligence company report on September 11. During a regular press briefing, Mao stated that China remains committed to the responsible development of AI, while firmly opposing distortions of

According to Beating, OpenAI has opened the GPT-Live-1 API, the real-time voice model powering ChatGPT. The model can simultaneously listen and speak, understanding speech directly and handling natural conversation, interruptions, pauses, and background noise. For complex reasoning or tool use, GPT-

According to Nasdaq's 2026 Global Voice of the Issuer Study conducted with Escalent, private markets ranked highest in client demand across all surveyed regions, with 82% of respondents in the Americas, 71% in Asia-Pacific, and 67% in Europe citing private markets among their top five areas. The sur

According to FLOP Labs announcement, the AI inference network project backed by Arthur Hayes launched a "14-Line Poetry Chain Challenge" on Sept 11 at 8 PM Beijing time through Sept 18 at 8 PM. AI Agents can form teams of 4-8 members to collaboratively create fourteen-line poems by chaining words.

According to a threat intelligence report released by Anthropic on Sept 11, Moonshot AI (developer of the Kimi model series) redirected user requests to Claude models without processing them through Kimi, using a network of 5,380 fraudulent accounts. From May to July, distillation interactions

Anthropic identified China- and Russia-linked actors using its Claude AI models for cyberattacks, surveillance and weapons-related work in activity occurring between December 2025 and August 2026. The findings appear in Anthropic's September threat-intelligence report, which documents seven

According to Anthropic's September threat-intelligence report covering December 2025 through August 2026, the company identified China- and Russia-linked actors among groups attempting to use Claude AI models for cyberattacks, surveillance and weapons-related work. The incidents span seven

According to its official press release on September 11, Salesforce introduced Enterprise AI Harness, described as a trusted execution foundation for enterprise AI combining six capabilities: context, agency, action, governance, security, and models through shared architecture. The company also

According to Tibo, head of OpenAI Codex, the company will discontinue its code model GPT-5.3-Codex-Spark next week due to declining usage and the availability of higher-performing alternatives, Tibo announced on X on

According to Business Insider, Anthropic posted and subsequently deleted a job listing on September 9 for a "Mega Account Executive, Meta" role with a salary range of $380,000 to $450,000. The unusual move of naming a specific client in a job title drew attention amid ongoing tensions

According to Securities Times, domestic AI chip maker Xunyuan Technology (688801) surged over 200% on its first trading day on the Shanghai Stock Exchange's STAR board today, rising from its initial public offering price of 142.18 yuan per share to around 400 yuan per share. Investors who secured on

According to BlockBeats, on September 11, crypto KOL @StandartXBT discovered through analysis of Long.xyz factory code that the Robinhood chain meme coin issuance platform may support tokenization of OpenAI and Anthropic on 1x leveraged pools, with tokens backed by Lighter. However, as of

According to PANews, on September 11, Xiaomi released and open-sourced Xiaomi-CocktailASR-1, an industrial-grade target speaker speech recognition model. The model addresses the cocktail party problem by

D.A. Davidson analyst Gil Luria stated that Palantir Technologies could be among the best-positioned companies to benefit from AI regardless of whether the current AI boom continues or cools, speaking on Jon Erlichman's Ticker Tank podcast. Luria called Palantir "the best company in the world,"

According to Bloomberg and Reuters, OpenAI CEO Sam Altman disclosed this week during a company-wide meeting that OpenAI is considering slowing down advanced artificial intelligence development. Altman indicated the company may coordinate with other AI labs to address mounting safety concerns,

South Korean President Lee Jae-myung will host the first Korea-Central Asia Summit on the 16th in Seoul, bringing together leaders from Kazakhstan, Uzbekistan, Turkmenistan, Tajikistan, and Kyrgyzstan for the government's first multilateral presidential meeting with the region. The summit aims to

According to PANews, Alipay announced the launch of AI wallet agents on September 11, introducing three major AI collection capabilities: VibePay, SkillPay, and MachinePay. The company also unveiled new use cases including AI-powered periodic purchasing, AI-assisted snatching, and AI-assisted

According to Artificial Analysis' independent testing, DeepSeek V4.1 Flash scored 40 points on the comprehensive Intelligence Index, surpassing its own V4 Pro 0813 at 36 points but trailing Kimi K3 at 44 points and GLM-5.3 at 45 points. The model shows significant cost advantages: V4.1 Flash

According to Anthropic, on September 10, the company detected and disrupted unauthorized efforts by Chinese AI labs—including Alibaba, Moonshot AI, and DeepSeek—to train models using Claude outputs. Anthropic said the Alibaba-linked operation was the largest distillation campaign it measured, with m