Microsoft says its new multi-model agentic security system found 16 Windows vulnerabilities
Microsoft’s Security Blog describes MDASH, a harness orchestrating 100+ AI agents across multiple models to discover and prove exploitable bugs.
Startups
Wirestock raises $23M to supply creative multi-modal data to AI labs
The company says it now supplies datasets across images, video, and 3D content to large model makers.
Model of the Week
WEEK 20 · MAY 2026
Claude Opus 4.7
Anthropic · LLM Chat · Frontier Model
Anthropic’s Claude Opus 4.7 is a frontier model aimed at complex technical work and agentic coding, with API availability and prompt caching / batch discounts.
$5 / $25 per 1M tokens (API)
Prompt caching (up to 90% savings)
Available via Claude API + major clouds
Explore all models
252 models · 18 categories
Latest
View all →
U.S. bank discloses security lapse after sharing customer data with AI app
Community Bank said an unauthorized AI app led to exposure of customers’ personal data, prompting disclosure after it detected the issue.
Google DeepMind demos an AI-enabled pointer meant to bring Gemini into everyday UI
DeepMind outlines interaction principles and shows experimental demos where a pointer can select on-screen elements and invoke Gemini without switching contexts.
AI-assisted hacking is already here, Google warns
Google says it may have found the first known case of cybercriminals using AI to identify and exploit an undisclosed zero-day flaw.
Editor's Picks
7-day highlights
This Week in AI & Tech
AI & Models
All AI & Models →
AI
OpenAI launches new voice intelligence features in its API
OpenAI added new Realtime API capabilities for voice conversations, live transcription, and translation, including a new GPT‑Realtime‑2 model.
AI
Khosla-backed robotics startup Genesis AI goes full-stack, demo shows
The $105M-seed startup unveiled its first robotics foundation model, GENE-26.5, alongside its own human-sized robotic hands and data-capture hardware to speed training and evaluation.
AI
Researchers say they ‘gaslit’ Claude into giving explosive-making instructions
Security researchers at Mindgard told The Verge they used flattery and social manipulation to get Anthropic’s Claude (Sonnet 4.x) to produce restricted content without directly asking for it.
AI
OpenAI releases GPT-5.5 Instant, a new default model for ChatGPT
OpenAI says the new default ChatGPT model reduces hallucinations in sensitive domains while keeping low latency, and it’s also available via the API as “chat-latest.”
Big Tech
All Big Tech →
Big Tech
Google shuts down Project Mariner
The experimental tool that could complete tasks across the web has been discontinued, with pieces folded into other Google offerings.
Big Tech
Chrome's AI features may be hogging 4GB of your computer storage
Some Chrome on-device AI features can trigger a multi‑gigabyte Gemini Nano model download, surprising users on low-storage machines.
Big Tech
Google Search rolls out new AI Mode and AI Overviews link and context features
Google says it’s updating AI Mode and AI Overviews to surface more links inline, highlight subscription content, and add link previews and community perspectives.
Big Tech
Google Home's Gemini AI can handle more complicated requests
Google is upgrading Google Home with Gemini 3.1 to better handle multi-step commands, alongside new automation and camera features.
Chips & Hardware
All Chips & Hardware →
Chips
AI boom pushes Samsung to a $1T valuation
TechCrunch says Samsung shares jumped on surging AI-driven demand for high-bandwidth memory (HBM), lifting the company past the $1 trillion mark.
Chips
NVIDIA and Corning announce long-term partnership to strengthen U.S. manufacturing for AI infrastructure
Corning plans major U.S. capacity expansions for optical connectivity and fiber needed in large AI clusters.
Chips
Intel stock more than doubles in April as AI-era CPU demand and turnaround hopes build
CNBC says Intel shares rose 114% in April — the best month in the chipmaker’s Nasdaq history — as investors rallied behind a turnaround story and growing AI-driven demand for CPUs and data-center capacity.
Chips
Google unveils new TPUs positioned for the ‘agentic era’
The company says the new TPU designs target faster frontier-model training and improved efficiency across the AI lifecycle.
Startups & Funding
All Startups & Funding →
Startups
There aren't enough rockets for space data centers. Cowboy Space raised $275 million to build them.
Cowboy Space says it raised a $275M Series B to pursue orbital data centers — and decided it needs to build its own rockets to scale.
Startups
Nebius agrees to buy Eigen AI in about $643M deal to speed up and cut costs of AI inference
Cloud provider Nebius is acquiring 20-person startup Eigen AI, which optimizes chip performance for running (not training) AI models.
Startups
Nvidia invests in Swedish legal AI startup Legora at $5.6B valuation
Nvidia’s NVentures joined a Series D extension for Legora, which sells agent-style tools for legal workflows and says it has surpassed $100M in ARR.
Startups
Shapes raises $8M to put AI characters into group chats
The app blends human group conversations with user-created AI ‘Shapes’ designed to participate alongside friends.
Policy & Regulation
All Policy & Regulation →
Policy
US to safety test new AI models from Google, Microsoft, and xAI
The US Commerce Department will evaluate certain new AI tools and features ahead of public release under voluntary agreements with the companies.
Policy
Google, Microsoft, and xAI agree to share early AI models with the U.S. for evaluation
The Wall Street Journal says the companies will provide early versions for government-led testing of capabilities and risks before public release.
Policy
The White House reportedly is working on an executive order about AI oversight and access
Report says officials are weighing AI oversight measures and priority government access to frontier models.
Policy
Google, Microsoft, and xAI will allow the US government to review their new AI models
The Commerce Department’s CAISI says it will conduct pre-deployment evaluations and targeted research with major AI labs to better understand advanced model capabilities.
Analysis & Deep Dives
All Analysis & Deep Dives →
Analysis
Google's AI architect lived rent-free in Elon Musk's head
A look at how Demis Hassabis and Google DeepMind shaped early OpenAI anxieties and strategy, via documents and testimony from the Musk–Altman trial.
Analysis
AI digs up decades of code debt. Patch up.
The UK’s National Cyber Security Centre warns AI-driven bug discovery could trigger a wave of vulnerabilities that overwhelms patch capacity.
Analysis
WSJ: Big Tech’s AI spending spree is getting expensive — and investors want payback
The Wall Street Journal argues that Microsoft, Amazon, Meta, and Alphabet are ramping capital spending for AI fast enough that depreciation is starting to bite, raising pressure to translate AI demand into durable profit.
Analysis
AI puts one fifth of London jobs at risk
A City Hall report estimates more than a million London jobs are significantly exposed to AI-driven task automation.