1. U.S. Government Orders Global Shutdown of Anthropic’s Top AI Models
The U.S. government issued an emergency directive compelling Anthropic to permanently disable access to its two most powerful models, Fable and Mythos, citing national security risks. The move triggered global market volatility, sparked ethical debates on AI sovereignty, and led to an unprecedented “AI recall” — the first of its kind in industry history. Anthropic called the decision a “serious misunderstanding” of model capabilities.
2. Alibaba’s AI-Powered Alipay Enters Inner Testing Phase
Ant Group quietly launched the beta version of its AI-native Alipay interface, dubbed “Abao.” Users can now navigate the entire financial ecosystem — from ordering food to managing investments — via natural language dialogue. The system integrates real-time payment, credit scoring, and personalized financial advice in a single conversational flow, signaling a paradigm shift in super-app design.
3. Ideal Auto Unveils “Mach M100”: World’s First Data-Flow AI Chip for Vehicles
Ideal Auto revealed the Mach M100, a proprietary automotive-grade AI chip built on a revolutionary data-flow architecture. It achieves 82%+ operational efficiency, outperforming NVIDIA’s desktop-class supercomputers in real-time sensor fusion and decision latency. The chip will debut in the 2027 Ideal L9, marking the first time a Chinese automaker leads global automotive AI hardware innovation.
4. Chinese AI Models Dominate Global Usage: DeepSeek-V4-Flash Tops Charts
From June 8–14, 2026, Chinese large models accounted for 18.42 trillion tokens in global usage — 41% of the worldwide total. DeepSeek-V4-Flash led with 4.41 trillion tokens, marking its fourth consecutive week at #1. The top four models by usage were all Chinese: DeepSeek-V4-Flash, MiniMax M3, Tencent Hy3 Preview, and Qwen2.5. U.S. models collectively trailed at 5.72 trillion tokens.
5. Microsoft Unveils MAI Family at Build 2026: A New Era of Reasoning Models
At Microsoft Build, the company launched the MAI (Microsoft AI) series:
- MAI-Thinking-1: 35B active parameters, MoE, 256K context — scored 97.0 on AIME 2025.
- MAI-Code-1-Flash: 5B parameters, optimized for VS Code and GitHub Copilot CLI — outperformed Claude Haiku 4.5 on SWE-Bench Pro (51.2 vs 35.2).
- MAI-Image-2.5: Ranked #3 on text-to-image leaderboards.
- MAI-Voice-2 and MAI-Transcribe-1.5: Support 15 and 43 languages respectively.
All models are in private preview with no public pricing yet.
6. Waymo Acquires Apple’s 5,500-Acre Autonomous Test Facility for $220M
Waymo finalized the purchase of Apple’s former autonomous driving test site in Arizona — a 5,500-acre facility with simulated urban, highway, and weather environments. The acquisition expands Waymo’s testing capacity by 40% and accelerates its path to full autonomy. Apple confirmed it is shifting focus to AI-driven sensor fusion R&D, not vehicle deployment.
7. Apple’s New Siri Shows Major Leap — But Still Lags Behind Competitors
According to Bloomberg’s Mark Gurman, Apple’s updated Siri, powered by a custom on-device LLM, now understands context, handles multi-turn requests, and integrates with third-party apps seamlessly. However, its performance remains equivalent to leading models from six months ago. Analysts note Apple prioritizes privacy over raw capability, creating a “capability gap” in the consumer AI race.
8. Google in Talks with Samsung to Diversify AI Chip Supply Chain
Google is in advanced negotiations with Samsung to co-develop and manufacture next-generation AI accelerators, reducing reliance on TSMC. The partnership aims to produce 5nm-class AI chips with custom memory architectures optimized for Gemini 3.5 Pro. If finalized, this would mark Samsung’s first major entry into the AI silicon market beyond memory and display.
9. China Achieves 99.99% Purity in Silicon-28 Isotope Production
Chinese researchers at the Shanghai Institute of Applied Physics successfully scaled production of ultra-pure Silicon-28 (isotopic abundance >99.99%), a critical material for quantum computing and next-gen semiconductors. This breakthrough eliminates foreign dependency on enriched silicon and enables domestic production of fault-tolerant qubits and high-efficiency AI chips.
10. AI in Sports: World Cup Introduces RefCam — AI-Powered Offside & 3D Line Detection
For the first time in World Cup history, AI-powered RefCam systems — using 3D pose estimation, millimeter-accurate ball tracking, and real-time player limb mapping — made all offside decisions. The system processed 12,000+ frames per second across 64 matches, reducing human error to near zero. Broadcasters now offer “AI View” overlays showing decision logic in real time.