✨ AI NEWS
A daily briefing of the AI stories that matter.
🤖 Today in AI

Jensen Huang's NVIDIA CEO: Private Ties Allegations, China Strategy Risks, and Expansion of Korea's Physical AI Ecosystem
Jensen Huang, CEO of NVIDIA, has acknowledged a personal relationship with an individual under suspicion of being a Chinese spy, and as a company-wide investigation proceeds, NVIDIA’s lobbying strategy to ease U.S. export controls on semiconductors to China and its plans to enter the Chinese AI market are facing significant scrutiny. Meanwhile, NVIDIA held a…

Google's Acquisition of Crusoe Wyoming Data Center and the Accelerating Frontier AI Agent Ecosystem Amid Infrastructure Turmoil
Google’s sudden acquisition of Chris Woychowski’s Wyoming data center project, which had been facing difficulties, marks a major expansion of its large-scale AI infrastructure. Alongside this, OpenAI announced the GPT-6 Cyber cybersecurity model, GitHub Copilot expanded its agent management capabilities, and Meta unveiled its collaborative space called Space…

Anthropic's Pre-IPO Signals Stagnant AI Spending... Accelerating Practical Validation of Efficiency-Focused Agents
As Anthropic’s IPO looms, data showing that while corporate AI adoption is rising, actual spending remains stagnant has cast doubt on investors’ high-growth expectations. At the same time, the industry is shifting its focus to practical execution and cost efficiency, ranging from routine wet-lab evaluations to robot learning, simulation development, and exte…

OpenAI Unveils 'GPT-6 Cyber,' a Specialized Cybersecurity Model... Setting New Standards for Digital Infrastructure Defense and Alignment Agents
OpenAI unveiled its alignment-focused security model, "GPT-6 Cyber," specialized in analyzing and mitigating vulnerabilities in digital infrastructure, directly addressing the rigorous verification phase surrounding the infrastructure protection capabilities of AI research institutes.

Deokdeokgo Targets Educational Sector Amid Google AI Restrictions; OpenAI Acquires Codex and Accelerates GitHub Copilot Support
As Google faces backlash for its decision to halt search service provision to school districts that restrict AI features, DuckDuckGo has launched a campaign to attract schools by highlighting its option to disable AI. Meanwhile, OpenAI announced the acquisition of Pierre Computer Company to enhance its Codex capabilities and launched GPT-5.6 Cyber on AWS Gov…

OpenAI Launches AI-Powered Internal Security Sprint... Expanding AI as Essential Infrastructure and Industry-Specific Autonomous Agents
OpenAI has launched a sprint to strengthen its internal security framework through collaboration with Cathedral Systems. Meanwhile, Google announced its policy for integrating AI into educational search, GitHub Copilot received JetBrains Air support, and NVIDIA expanded its semiconductor fabs and smart city physical-agent AI initiatives, reflecting the broad…

Vercel Labs' agent-desktop: Automating AI Desktop Manipulation via Accessibility Tree and Unique References
agent-desktop, released by Vercel Labs, is a CLI and C library for AI agents that manipulate desktop applications by reading the OS accessibility tree. It stably performs the 'observe-manipulate-reobserve' loop by assigning unique references to UI elements instead of relying on screen pixel matching, and supports macOS, Windows, and Linux environments. With…

Claude-Mem: A Memory Compression System for Preserving Context Across Sessions in Claude Code
The open-source project 'Claude-Mem' is a continuous memory compression system that helps maintain continuity of project knowledge even after Claude Code sessions are terminated or reconnected. It automatically records tool usage observations, generates semantic summaries, and passes them to subsequent sessions, supporting various environments such as Grok B…

42AI Unveils Stealth Mode After Raising $62.5 Million for Human Behavior Prediction World Model
AI startup 42AI has emerged from stealth mode after securing $62.5 million in funding, leveraging its world model technology capable of predicting human behavior even in new environments. Meanwhile, as AI operational costs surge, Elon Musk is tightening usage limits for Grok Bot, and OpenAI is providing support to stabilize the power grid for its computing c…

Orca: A Tool for Viewing the Work of Multiple Coding Agents in One Place
Orca is an open-source project designed to run multiple CLI coding agents in separate Git worktrees and monitor their status and changes from a single location. This article introduces parallel processing, remote access, and mobile verification features documented in the public README, and outlines criteria for applying Orca to research-oriented code managem…