All Updates
Page 449 of 862
March 20, 2026
Meta Agent Leaks Secrets, Zuck Sounds Alarm
Meta's AI Agent malfunctioned, leading to a data leak. Mark Zuckerberg issued a top-level emergency alert. The incident highlights AI risks even for industry leaders.
Huang: Industrials Become Robots, Nvidia Drops AI Stack
Jensen Huang predicts every industrial firm will transform into a robotics company. Nvidia unveiled a comprehensive physical AI infrastructure suite. This positions Nvidia as a leader in embodied AI.
Super Micro Faces US Charges Over AI Chip Exports
US prosecutors have charged three individuals linked to Silicon Valley server maker Super Micro, including a co-founder, with violating export laws. The men allegedly diverted servers containing Nvidia AI chips to China. This case highlights escalating US scrutiny on AI hardware exports.
China Launches Largest Humanoid Robot Base
Beijing's Shijingshan district opened China's largest robot training base, over 10,000 sqm across three phases. It emphasizes tactile and multimodal perception training. Robots detect forces as low as 0.01N, equivalent to a hair strand's pressure.
Xiyu Tech Launches MiniMax-M2.7 Model
Xiyu Technology released the MiniMax-M2.7 AI model in today's Pai Morning Report. The roundup also highlights Xiaomi's launch of multiple new products, Google Stitch's vibe design introduction, and Philips' new Moving Sound series.
Microsoft Sets Win11 26H1 Support Timeline
Microsoft confirmed Windows 11 26H1 support starts February 10, 2026: Home/Pro for 24 months until March 14, 2028; Enterprise/Edu for 36 months. Spring release adapts to AI+PC hardware ecosystem. Existing devices get 26H2 later; 24H2 support ends October 2026.
Bezos Plans $100B AI Manufacturing Fund
Amazon founder Jeff Bezos is reportedly raising $100 billion for a fund to acquire manufacturing companies and integrate AI technologies. He has held meetings with major asset managers during trips to the Middle East and Singapore. Investor documents confirm ongoing fundraising efforts.
Qwen Code v0.13.0 Nightly: Hooks, Arena, Concurrent Tools
Qwen Code's v0.13.0 nightly release introduces hooks extension, concurrent task tool execution for better performance, and an agent collaboration arena with multi-model support. It adds system prompt customization in SDK/CLI, UI token usage display, and VSCode enhancements like Tab fill-only completions. Fixes cover duplicate chunks, error handling, and model token limits.
OpenAI Plans Desktop Superapp
OpenAI is developing a desktop superapp merging ChatGPT, Codex AI coding app, and Atlas browser. The move aims to simplify products and reduce fragmentation slowing quality improvements, per CEO Fidji Simo's memo. This comes amid competition from Anthropic.
UgREEN AI NAS Hype Fades
UgREEN's stock surged on AI NAS hype with OpenClaw support but crashed back as transformation lags expectations. NAS revenue booms 315% but low margins and giant competition from Xiaomi, Huawei loom. Questions sustainability of first-mover edge.
March 19, 2026
Tencent Stock Drops on AI Investment Plans
Tencent shares plunged over 6% despite beating earnings expectations, driven by signals of ramped-up AI investments and potential buyback reductions. The strategy aims to counter rivals like ByteDance, Alibaba, and Baidu. Additional AI notes include HSBC eyeing 20K AI-driven layoffs and BlackRock CEO warning of record grad unemployment.
CISA Urges Securing Intune Post-Iran Hack
CISA warns IT leaders to harden Microsoft Intune configurations after pro-Iranian Handala group hacked Stryker. Key measures include phishing-resistant MFA, least privilege access, and multi-admin approvals. Advice extends to all endpoint management systems.
Lime's AI Vision Calls Out Sidewalk Riders
Lime Vision employs a front-mounted camera and trained AI model on electric scooters to detect improper riding locations like sidewalks. The system aims to boost safety for riders and pedestrians. It will debut nationally in Seattle in June.
Global Memory Surge: 32GB Hits 3800 Yuan
Memory prices worldwide have exploded, with 32GB strips jumping from 800 to 3800 yuan—a 375% rise. Kingston 16GB basics rose from 200 to over 800 yuan since last spring. PC and gaming costs soar, driving demand for second-hand alternatives.
Humanoid Robots Deploy in Auto Factories
Major carmakers like Hyundai, Toyota, BMW, Xiaomi, XPeng, and Tesla are deploying humanoid robots such as Atlas, Digit, AEON, CyberOne, IRON, and Optimus in factories for tasks like sorting, handling, and assembly starting 2026. These robots achieve high success rates and efficiency gains, targeting total assembly lines next. IDTechEx predicts 1.6M units in auto industry by 2035.
Tencent's Aggressive AI Pivot with Longxia Apps
Tencent launched over 10 Longxia AI apps in two weeks, signaling strong AI commitment amid robust Q4 2025 earnings. AI drives ad revenue growth to 19% with 3x click rates, game revenues up 22% via AI NPCs, and WeChat integrations like QClaw for PC automation. Pony Ma views AI as a century-scale revolution reshaping Tencent's ecosystem.
Super Micro Board Member Indicted for AI Tech Theft
Three people associated with Super Micro Computer Inc., including a board member, were indicted for allegedly plotting to divert billions of dollars in AI technology to China. The California-based company disclosed the indictments. This raises concerns over supply chain security in AI hardware.
NVIDIA CEO Urges No AI Panic from Tech Leaders
NVIDIA CEO Jensen Huang advised tech leaders to avoid scaring the public about AI. Responding to the Anthropic-Pentagon dispute, he said warnings are good but fear-mongering is harmful. He emphasized AI's critical importance during a GTC panel.
Pentagon Warns of Anthropic Foreign Staff Risks
US DoD filed a court document highlighting Anthropic's heavy reliance on foreign employees, including Chinese nationals, for large model R&D and support. This hiring structure raises fresh national security concerns. The warning ties to ongoing disputes.
TorchSpec Enables Scaled Speculative Decoding Training
PyTorch introduces TorchSpec, a new system for training speculative decoding at scale amid rapid growth in large language models. Frontier models like Kimi K2.5, GLM 5, and Qwen 3.5 highlight the need for efficient training techniques. TorchSpec aims to support these massive models effectively.