Vidu AI, founded by Tsinghua professor Zhu Jun, has completed a nearly 2 billion RMB Series B round led by Alibaba Cloud, raising its valuation to 12 billion RMB and officially entering the unicorn ranks. The company's core product, the Vidu video generation large model, has undergone three iterations and surpassed models like Runway and Sora on the VBench leaderboard, ranking first in China and second globally in international benchmarks. It has raised a total of 2.6 billion RMB over the past two months, completing seven funding rounds within three years, with its valuation growing more than 16-fold. Alibaba has led investment twice, combining its AI video model HappyHorse-1.0 and newly established ATH business group to accelerate布局 in the AI video generation sector. The company expects user base and revenue to grow over 10x in 2025, with Goldman Sachs projecting the industry to expand from $3 billion in 2025 to $29 billion by 2030.

Anthropic has released the public beta of Claude for Word, integrating the Claude AI into Microsoft Word as a sidebar plugin, available to Team and Enterprise users. All AI-generated edits appear in Word’s native track-changes mode, preserving formatting and edit history. The key innovation is cross-application context sharing, allowing联动 with Claude for Excel and PowerPoint for multi-document collaborative processing, ideal for professional domains such as finance and law. Its "advisor-executor" architecture enables high-intelligence models to plan while efficient models draft. Anthropic warns users to guard against prompt injection attacks, with all patch suggestions requiring human approval. This move marks Anthropic’s differentiated competition with Microsoft 365 Copilot in the enterprise AI productivity space.

On April 11, Anthropic launched Claude Code Security, now in limited preview for enterprise and team customers. Leveraging AI for multi-stage reasoning and validation, the feature identifies logical flaws and data leakage risks that traditional static analysis tools often miss, providing severity ratings and remediation suggestions—though all patches require manual approval. Part of the $104 million Project Glasswing initiative, it collaborates with over 40 organizations including AWS, Apple, Google, Microsoft, and NVIDIA to advance AI-powered security defense. Anthropic is offering $100 million in usage credits and a $4 million donation to support open-source security projects. Currently, access is restricted to advanced users.

The Ministry of Education, National Development and Reform Commission, Ministry of Industry and Information Technology, Ministry of Science and Technology, and National Data Bureau jointly issued the "AI+Education" Action Plan, outlining four key tasks during the 15th Five-Year Plan period: cultivating AI talent and enhancing literacy, deep integration of AI and education, infrastructure development, and open ecosystem creation. The plan promotes AI curriculum across all levels from K–12 to lifelong learning, proposes inclusion in teacher certification exams, supports intelligent learning companions and teaching systems, and aims to build a national intelligent computing platform for education. It also mandates improved AI risk预警 and emergency mechanisms to combat deepfakes, fraud, and academic misconduct. Institutions like Tsinghua and Fudan Universities have already initiated related practices.

A recent Morgan Stanley report highlights that top-tier large models are undergoing "non-linear capability leaps," with AI compute demand now systematically exceeding supply. From January to March 2026, global weekly token usage spiked from 6.4 trillion to 22.7 trillion—an approximately 250% increase. Pre-compute demand growth is expected to be three times NVIDIA’s compound annual supply growth rate. On energy, U.S. data centers face an estimated 55GW power deficit between 2025 and 2028, with $18 billion in projects canceled and $46 billion delayed. Labor market impacts are emerging: in the five most AI-affected industries, 11% of positions have been eliminated, and 90% of occupations will experience varying degrees of impact. Cost savings from AI automation could amount to over 25% of S&P 500 pre-tax profits in 2026.

Cisco announced the acquisition of AI observability startup Galileo Technologies to enhance Splunk’s capabilities in monitoring and trust for AI agents. Galileo specializes in real-time monitoring, bias, and hallucination detection for multi-agent AI systems, already adopted by companies like Comcast and HP; the deal is expected to close in Cisco’s FY2026 Q4. Simultaneously, Cisco is reportedly acquiring Israeli AI agent security startup Astrix Security for $250–350 million. Astrix focuses on AI agent identity protection and anomaly detection, capable of automatically identifying non-human identities and vulnerabilities in MCP servers. Together with Cisco AI Defense, these acquisitions form a three-layer security architecture, signaling a shift toward comprehensive governance in enterprise AI security.

Three key figures who led OpenAI’s multi-billion-dollar 'Stargate' data center project—Peter Hoeschele, Shamez Hemani, and Anuj Saharan—are planning to join Meta. Stargate is OpenAI’s $500 billion AI infrastructure initiative with Oracle and SoftBank, but OpenAI has paused its UK project and canceled Texas expansion leases to control spending ahead of IPO. Meanwhile, Meta is projected to spend up to $135 billion in capital expenditures this year, heavily investing in AI infrastructure, recently launching the Muse Spark model through its Super Intelligent Lab. This talent shift reflects evolving dynamics in the AI infrastructure race.

The Ministry of Industry and Information Technology (MIIT) announced it will release a series of high-value "AI+" scenarios in manufacturing to promote deeper AI integration. In 2025, China’s AI core industry scale exceeded 1.2 trillion RMB, with over 6,200 enterprises and AI technology adoption exceeding 30% among designated manufacturing firms. Next steps include the 'Model-Data Resonance' initiative and humanoid robot real-world training programs, advancing standards development and ethical governance, and establishing the China-BRICS Artificial Intelligence Development and Cooperation Center. The China Academy of Information and Communications Technology (CAICT) notes AI is entering the era of intelligence-native systems, with foundational large models evolving into full-stack intelligent agents, though data privacy and privilege abuse risks require concurrent governance.

In early April, security researchers disclosed over 30 critical vulnerabilities in major AI coding assistants, impacting GitHub Copilot, Amazon CodeWhisperer, and Tabnine, potentially leading to data exfiltration and remote code execution. The most severe flaws include Copilot’s CVE-2026-4512 and CodeWhisperer’s buffer overflow vulnerability CVE-2026-7891. Vendors issued emergency patches between April 9 and 11. The study found 68% of AI coding tools lack threat modeling during development, prompting experts to call for a standardized security framework, the 'AI Security Baseline' (ASB-2026), though only 15% of major vendors currently participate. Developers are advised to update tools immediately and disable AI suggestions for sensitive code.

AI-powered end-to-end patent lifecycle management platform Patlytics has raised $40 million in Series B funding led by SignalFire, bringing total funding to approximately $65 million. Founded in 2024, the platform is used by over 40% of Am Law 100 firms and enterprises such as Rivian, Xerox, and Canon for patent drafting, infringement analysis, and portfolio management. Customer feedback indicates an 80% increase in project completion speed, saving over $30,000 per claim chart and up to 15 hours per patent application. The company plans to focus on AI collaboration tools for law firms and corporate IP teams, AI agents capable of autonomous research and drafting, and enhanced support for chemical and biological patents.