AI News: April 9, 2026

1. Zhipu Open-Sources GLM-5.1: 744B MoE Agentic Coding Model Tops SWE-Bench Pro

Zhipu (Z.ai). Zhipu released GLM-5.1, a 744-billion-parameter MoE model (40B active) under MIT license, purpose-built for long-horizon agentic coding tasks up to 8 hours autonomous. It scored 58.4 on SWE-Bench Pro, topping GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro on that benchmark. Features a 200K context window and 131K max output, available on Hugging Face and compatible with Claude Code, OpenClaw, and Cline. Source

2. MATCH Act Would Ban DUV Lithography Sales to Chinese Chipmakers

Policy. A bipartisan MATCH Act (Multilateral Alignment of Technology Controls on Hardware) introduced in Congress would ban ASML’s DUV lithography machines and their maintenance to SMIC, Hua Hong, Huawei, CXMT, and YMTC, significantly tightening U.S. chip export controls beyond existing EUV restrictions. ASML shares fell 2.6% on the news. Bank of America estimates a full ban could reduce ASML revenues 14-15% and EBIT 16-17%. Source

3. Volkswagen and Uber Begin Testing Autonomous ID. Buzz Microbuses in Los Angeles

Autonomous Vehicles. Volkswagen subsidiary MOIA America and Uber started testing approximately 10 autonomous ID. Buzz microbuses in Los Angeles with a 27-sensor suite (13 cameras, 9 LiDAR, 5 radar) and Mobileye computing. Commercial rides with safety operators are expected by end of 2026 with fully driverless operation in 2027. LA is the first city in a planned multi-city rollout. Source

4. AI Medical Scribes Acknowledged as Increasing Healthcare Costs

Healthcare. STAT News reports that insurers and hospitals privately agree AI medical scribes are increasing healthcare costs by enabling more thorough documentation that codes visits at higher complexity levels. AI scribes also increased weekly visit volume by 1.7% and revenue by $167/month per clinician. Health economists warn of an “AI coding arms race” between AI-maximized billing and insurer algorithms minimizing payments. Source

5. Anumana Receives First-Ever FDA Clearance for ECG-AI Cardiac Amyloidosis Detection

FDA. Anumana, co-founded by Mayo Clinic and nference, received the first and only FDA clearance for an ECG-AI algorithm detecting cardiac amyloidosis using standard 12-lead ECGs. The algorithm achieved 78.9% sensitivity and 91.2% specificity for a rare, serious, and underdiagnosed heart condition that has historically required invasive biopsy or specialized imaging for diagnosis. Source

6. Etiometry Gets First-Ever FDA Clearance for AI Cardiogenic Shock Classification

FDA. Boston-based Etiometry received the first-ever FDA clearance for software that automates classification and tracking of cardiogenic shock. The AI-powered platform continuously analyzes patient data from medical devices and EHR systems to provide real-time clinician insights for this medical emergency, reducing time to intervention. Source

7. Eclipse Ventures Raises $1.3B for Physical AI, Robotics, and Infrastructure

Venture Capital. Eclipse Ventures, backer of Cerebras and Wayve, raised $1.3 billion across two funds (Fund VI at $720M, Early Growth Fund III at $591M), bringing AUM to approximately $10 billion. The firm focuses on physical AI across transportation, energy, defense, and manufacturing, with a unique hybrid model that includes incubating startups internally. Source

8. Hermeus Raises $350M at $1B Valuation for Autonomous Hypersonic Fighters

Defense AI. Hermeus hit unicorn status with $350M ($200M equity + $150M debt) led by Khosla Ventures, with Founders Fund, In-Q-Tel, and RTX Ventures participating. The company builds autonomous hypersonic aircraft using a modified Pratt & Whitney F100 engine and has completed two successful test flights. AI-driven autonomy is central to the unmanned fighter program. Source

9. D-Robotics Raises $150M Series B2 for Embodied AI Robotics Platform

Venture Capital. China-based D-Robotics closed a $150M Series B2 round (total Series B: $270M), backed by Didi Global, Prosperity7, and GL Ventures. The company builds embodied AI computing platforms (HoloBrain, HoloMotion) for robotics, with 2025 shipments surging 180% year-over-year and customer base growing 200%. Source

10. Databricks Co-Founder Matei Zaharia Wins ACM Computing Prize

Industry. Matei Zaharia, co-founder of Databricks and creator of Apache Spark, won the ACM Prize in Computing, the field’s top honor. Now working on AI for research at Databricks, Zaharia argues that AGI is misunderstood and that current systems already qualify if the definition is properly framed. Source

11. Atlassian Launches Visual AI Tools and Third-Party Agents in Confluence

Enterprise AI. Confluence users can now create visual assets within the software, and Atlassian is adding third-party agent integrations from Lovable, Replit, and Gamma. The update expands Atlassian’s AI footprint beyond text-based productivity into visual content creation and multi-agent workflows within its collaboration platform. Source

12. MinerU-Diffusion: Open-Source OCR Model Replaces Autoregressive Decoding with Diffusion

Open Source. Researchers from Shanghai AI Lab and Peking University introduced MinerU-Diffusion, a 2.5-billion-parameter open-source model that replaces traditional autoregressive OCR decoding with parallel diffusion denoising. It delivers up to 3.26x faster throughput with near-parity accuracy, arguing that document reading should follow visual structure rather than left-to-right token generation. Released under MIT license. Source

13. Study Finds 1 in 4 AI Chatbot Quotes Comes From Journalism

Research. A Muckrack analysis of 15 million AI citations from ChatGPT, Claude, and Gemini found that one in four source references traces back to journalism. Trade publications and specialist journalists benefit most from AI citation, while general news outlets rank lower. The study quantifies how heavily AI chatbots rely on journalistic content for their responses. Source

14. AI-Powered Abuse Ecosystem Documented Across 2.8 Million Telegram Messages

Safety. An analysis of 2.8 million Telegram messages in Italy and Spain documents how AI tools are fueling a monetized ecosystem built around non-consensual intimate imagery. The report details how nudifying bots, deepfakes, and automated archiving combine into an organized abuse pipeline that exploits AI-generated content at scale. Source

15. Poke Launches AI Agents Accessible via Text Message

Product Launch. Startup Poke launched a service that brings AI agents to everyday users via text message, handling tasks and automations without requiring apps, accounts, or technical knowledge. The product targets the consumer market where most AI agent tools remain too complex for non-technical users. Source

16. Astropad Workbench Reimagines Remote Desktop for AI Agent Monitoring

Developer Tools. Astropad launched Workbench, a tool that lets users remotely monitor and control AI agents running on Mac Minis from iPhone or iPad with low-latency streaming. It repurposes remote desktop technology specifically for the AI agent use case, providing a mobile control plane for autonomous workflows. Source

17. NIST Releases Concept Note for AI Risk Management in Critical Infrastructure

Policy. NIST released a concept note for a new AI Risk Management Framework Profile specifically targeting critical infrastructure operators. The profile will guide risk management practices for AI-enabled capabilities in critical infrastructure environments, extending NIST’s voluntary AI RMF 1.0 framework to address sector-specific deployment risks. Source