Tags: China

Google’s Threat Intelligence Group disclosed that a criminal hacking crew used an artificial‑intelligence model to locate a zero‑day flaw in a widely used open‑source system‑administration platform. The vulnerability could have bypassed two‑factor authentication and enabled a mass exploit across multiple organizations. Google intervened, alerted the software’s developers, and helped roll out a patch before the attack could be launched. The report, which does not identify the attackers, the software, or the AI model, also notes growing interest from state‑linked groups in AI‑assisted hacking tools. Ler mais

Chinese AI lab DeepSeek is in talks to raise its first venture‑capital round, aiming for a valuation that could climb from $20 billion to $45 billion. Founder Liang Wenfeng, who controls about 90% of the company, is turning to investors after rival firms began poaching key researchers. The round is expected to be led by the state‑backed China Integrated Circuit Industry Investment Fund, with cloud giants Tencent and Alibaba reportedly considering participation. DeepSeek’s efficient large‑language model runs on Huawei chips, positioning it as a strategic asset in Beijing’s push for AI independence. Ler mais

Meta is preparing to reverse its roughly $2 billion purchase of AI startup Manus after China’s National Development and Reform Commission issued a formal order canceling the transaction. The regulator gave the companies a few weeks to restore Manus’s Chinese assets to their pre‑acquisition state. The move forces Meta to write down a major AI investment and raises questions about how China will treat foreign‑owned tech firms with links to the country. Ler mais

China’s National Development and Reform Commission ordered Meta to unwind its $2 billion acquisition of Manus, an agentic AI firm founded by Chinese engineers and now based in Singapore. The regulator gave no public explanation, demanding the parties pull the deal and imposing exit bans on the startup’s founders. About 100 Manus staff have already moved into Meta’s Singapore offices, and the block threatens Meta’s push into AI agents while raising fresh concerns in Washington about Chinese‑linked technology investments. Ler mais

China’s decision to halt Meta’s purchase of AI firm Manus has thrown the deal into limbo, raising fresh uncertainty for both companies’ artificial‑intelligence plans. Manus founders Xiao Hong and Ji Yichao had already moved most of their staff to Meta’s Singapore office and stripped the startup of Chinese ties. The cancellation could prevent Manus from using Anthropic’s Claude models and marks a setback for Meta’s pivot toward AI after years of investing in the metaverse. Industry observers say the move underscores the growing tech rivalry between Washington and Beijing. Ler mais

DeepSeek, the Beijing‑based artificial‑intelligence startup, released a preview of its next‑generation V4 model on Friday, claiming it can match the performance of leading U.S. systems from Google, OpenAI and Anthropic. The open‑source model emphasizes coding capabilities and boasts compatibility with domestic Huawei hardware, marking a milestone for China’s chip industry. The announcement follows DeepSeek’s earlier R1 model, which drew criticism from U.S. officials over alleged use of banned Nvidia chips and accusations of misusing Anthropic’s Claude technology. Ler mais

President Donald Trump’s effort to fast‑track AI data center construction is hitting a wall. Developers say aggressive tariffs on Chinese imports have crippled the supply chain for essential power‑equipment components, pushing delivery times from months to years. With U.S. manufacturers unable to meet demand, nearly half of the year’s planned facilities face delay or cancellation, jeopardizing the administration’s goal of keeping America ahead in the artificial‑intelligence race. Ler mais

OpenClaw, a free open‑source AI agent created by Austrian programmer Peter Steinberger, has seen explosive uptake in China. The tool, which links to any large language model and automates tasks, attracted government subsidies, cheap domestic model access, and a workforce anxious about AI‑related job loss. In contrast, the United States and European Union have seen slower adoption due to higher model costs, privacy regulations, and a lack of coordinated support. While Chinese users benefit from lower running costs and institutional backing, concerns about security and data exposure persist across all markets. Ler mais

Alibaba has introduced a mobile app called JVS Claw that lets users install and run OpenClaw AI agents without coding. The launch follows Baidu’s similar offering and reflects intense competition among China’s leading AI firms to capture a broad consumer base. While adoption of OpenClaw continues to rise, authorities have expressed mixed signals: several local municipalities are providing subsidies to spur development, yet Beijing has barred state‑run enterprises from deploying the technology over cybersecurity worries. Experts warn that the open‑runtime model could expose users to data theft and malware. Ler mais

Researchers from Stanford and Princeton compared the responses of several Chinese and American large language models to politically sensitive questions. The study found that Chinese models refuse to answer a significantly larger share of these queries, provide shorter replies, and sometimes deliver inaccurate information. The authors suggest that manual fine‑tuning, rather than censored training data, drives much of this behavior. Additional work shows that extracting hidden instructions from Chinese models is difficult, highlighting the challenges of studying AI‑driven censorship in real time. Ler mais

Google’s Threat Tracker report reveals that hackers are conducting "distillation attacks" by flooding the Gemini AI model with more than 100,000 prompts to steal its underlying technology. The attempts appear to originate from actors in North Korea, Russia and China and are classified as model extraction attacks, where adversaries probe a mature machine‑learning system to replicate its capabilities. While Google says the activity does not threaten end users directly, it poses a serious risk to service providers and AI developers whose models could be copied and repurposed. The report highlights a growing wave of AI‑focused theft and underscores the need for stronger defenses in the rapidly evolving AI landscape. Ler mais

China has granted conditional approval for the import of Nvidia's high-end AI GPUs, signaling a strategic move to support its major internet firms while protecting domestic semiconductor interests. The approvals arrived during Nvidia CEO Jensen Huang's visit to Beijing, but strict licensing terms and restrictions on state-backed entities suggest a cautious approach. Chinese tech companies await future rounds of permission, and regulators are weighing the need for foreign chips against the goal of boosting local chip production. Ler mais

China’s top internet regulator, the Cyberspace Administration of China (CAC), requires every AI tool with public‑opinion or social‑mobilization capabilities to be filed in a public algorithm registry. The resulting database reveals thousands of generative‑AI and deep‑synthesis tools spread across major tech hubs such as Beijing, Shenzhen, Shanghai and Hangzhou, as well as emerging centers like Chongqing, Hefei and Guizhou. State‑linked firms account for a significant share of filings, while foreign companies represent a tiny fraction. Competition is diverse, with six “AI tigers” backed by Alibaba or Tencent, and a wave of startups applying AI to education, traditional medicine, carbon accounting, robotics and immersive entertainment. Many of these firms are already eyeing overseas markets, illustrating China’s growing global AI footprint. Ler mais

Researchers in China have unveiled photonic AI chips that reportedly outperform conventional Nvidia GPUs by up to 100 times on narrowly defined generative tasks. The hybrid ACCEL system combines optical and analog electronic components, while the all‑optical LightGen chip contains more than two million photonic neurons. Both platforms claim dramatic improvements in speed and energy efficiency for image‑related workloads, though they are targeted at specialized applications rather than general‑purpose computing. Ler mais

China's Cyberspace Administration has drafted comprehensive regulations aimed at curbing harmful behavior by AI chatbots. The proposal would apply to any AI service available in the country that simulates human conversation through text, images, audio or video. Key provisions require immediate human intervention when users mention suicide, mandate guardian contact information for minors and the elderly, and ban content that encourages self‑harm, violence, obscenity, gambling, crime or emotional manipulation. Experts say the rules could become the world’s most stringent framework for AI companions, addressing growing concerns about mental‑health impacts and misinformation. Ler mais

The United States is confronting a growing clash over how to regulate artificial intelligence, drawing parallels to the hands‑off approach of the early Internet era. While some lawmakers pushed the Telecommunications Act of 1996 to give the FCC oversight, modern efforts focus on preventing an AI arms race with China and addressing concerns about bias, misinformation, and job security. The White House has issued an executive order to block state‑level AI rules, arguing that fragmented regulation would hinder national competitiveness. Meanwhile, the EU has moved faster on user‑data protections, highlighting divergent global strategies. Ler mais

President Donald Trump announced that Nvidia has received White House approval to export its H200 AI GPU chips to "approved customers" in China and other nations. The deal includes a 25 percent U.S. cut of sales, higher than the 15 percent cut previously demanded. The H200 chips are an upgrade to the scaled‑down H20 models already allowed for export, though they remain less powerful than Nvidia's flagship Blackwell GPUs. Trump framed the policy as a boost for American jobs, manufacturing, and taxpayers, while noting that Nvidia’s U.S. customers are already moving forward with Blackwell chips. Senators have urged continued restrictions on China’s access to the most advanced AI chips. Ler mais

The U.S. Department of Commerce has authorized Nvidia to ship its H200 artificial‑intelligence chips to approved Chinese customers. The approval allows the company to sell chips that are roughly 18 months old, with the United States taking a 25% share of the sales. The decision comes amid bipartisan legislative efforts to block advanced AI chip exports and follows a series of policy shifts on U.S. chip sales to China. Lawmakers have introduced the Secure and Feasible Exports (SAFE) Chips Act, which would prohibit such exports for up to 30 months. Ler mais

Nvidia chief executive Jensen Huang cautioned that China can construct AI data centers and even hospitals far faster than the United States, citing the country's expansive energy resources and swift construction capabilities. At the same time, a report from OpenRouter and Andreessen Horowitz shows Chinese open‑source large language models now account for roughly 30% of global AI token usage, up from just over 1% a year earlier. While Huang affirmed Nvidia’s chip technology remains ahead of China, the rapid growth of Chinese AI models and the nation’s infrastructure advantages highlight an intensifying competitive landscape. Ler mais

Chinese startup DeepSeek has released two new AI models—DeepSeek‑V3.2 and DeepSeek‑V3.2‑Speciale—under an open-source license. The models claim performance comparable to GPT‑5 and Gemini 3 Pro on long‑form reasoning, tool use, and dense problem solving while offering a 128,000‑token context window and reduced computational cost through Sparse Attention. Their launch challenges the dominance of U.S. AI firms, sparks regulatory scrutiny in Europe, and raises questions about the future of AI accessibility and geopolitics. Ler mais