AI News
The latest developments, breakthroughs, and analysis from the world of artificial intelligence.
Google DeepMind Pays $90M for Contextual AI Talent in Antitrust Workaround
DeepMind acquires Contextual AI's CEO and 20+ researchers via licensing deal — Google's repeatable acquihire pattern to dodge US antitrust review.
Japan Airlines Begins Humanoid Robot Trial at Haneda Airport
JAL deploys Unitree humanoid robots at Tokyo's Haneda for baggage and cabin work — Japan's first airport humanoid trial, running through 2028.
Mistral Explores Custom AI Chips as Europe Buildout Accelerates
CEO Arthur Mensch confirms Mistral is studying its own silicon while a 13,800-GPU Paris site and a $1.2B Sweden investment come online.
OpenAI Brings Codex Computer Use to Windows With Mobile Steering
Codex can now click, type, and drive Windows apps in the foreground while you steer the work from your phone or a Mac.
OpenAI Drops Canvas From GPT-5.5, Bakes Writing and Coding Into Chat
GPT-5.5 Instant gets clearer responses and inline writing and coding blocks, while canvas disappears from Instant and Thinking modes.
OpenAI Sets Retirement Dates for o3 and GPT-4.5 in ChatGPT
GPT-4.5 sunsets June 27 and o3 retires August 26 as OpenAI clears its lineup for newer flagship models. Here's what users need to do.
Amazon's Custom Chips Hit $20B Run Rate, Trainium Sold Out
Andy Jassy puts Amazon's silicon business at a $20B run rate growing triple digits. Anthropic locked up 5 gigawatts of Trainium; demand still outpaces supply.
Anthropic's Glasswing Finds 10,000+ Software Flaws in One Month
Project Glasswing surfaced over 10,000 high-severity bugs in 30 days using Claude Mythos — and 99% remain unpatched, exposing a new patching crisis.
Anthropic Opens Seoul Office, Names KiYoung Choi to Run Korea
Korean Claude usage runs 3.5x ahead of population. Anthropic's new Seoul office and ex-Snowflake country head formalize a market that was already paying.
DuckDuckGo Installs Surge 30% as Users Reject Google's AI Search
DuckDuckGo's US iOS installs peaked at 69.9% week-over-week growth after Google replaced search results with AI agents. The opt-out market is real.
IBM Joins Anthropic's Project Glasswing, Brings Concert to the Fight
IBM joins Project Glasswing as the 50th+ partner. Concert plugs the remediation gap where Glasswing's 10,000-bug discovery rate is overwhelming defenders.
Meta Plans AI Pendant and 'Wearables for Work' Hardware Push
Meta is developing an AI pendant and an enterprise wearables line, building on its 2025 Limitless acquisition, with testing planned in 2027.
Micron Hits $1 Trillion Market Cap as AI Memory Demand Explodes
Micron joined the trillion-dollar club after UBS tripled its price target to $1,625, citing sold-out HBM capacity and structural AI memory shortages.
Microsoft Build 2026 Preview: Windows Becomes an Agent Platform
Microsoft's June 2 conference will turn Windows into an OS for AI agents, with a new Agent Store, Agent Framework APIs, and a multi-model Copilot.
Claude Opus 4.8 Lands With Dynamic Workflows and 3x Cheaper Fast Mode
Anthropic's new flagship arrives 41 days after Opus 4.7 with a workflow runtime that orchestrates up to 1,000 subagents and sharp honesty gains.
CNN Sues Perplexity Over 17,000 Stories in First TV-Network AI Case
CNN's federal copyright and trademark suit accuses Perplexity of scraping 17,000+ stories and falsely implying an affiliation with the network.
DeepSeek Makes 75% V4-Pro Price Cut Permanent, Reigniting AI Price War
DeepSeek's V4-Pro now permanently costs $0.435/M in and $0.87/M out — roughly 7x cheaper than Claude Opus and 9x cheaper than GPT-5.5.
Google Flips Core Search to Gemini 3.5 Flash, Making AI Mode the Default
Google's AI Mode now runs on Gemini 3.5 Flash by default, turning the global search box into a conversational interface for every user.
OpenAI Releases Frontier Governance Framework to Align With New AI Laws
OpenAI's new framework maps its safety practices to California's SB 53 and the EU AI Act — and signals where it expects regulators to look first.
OpenAI Launches Rosalind Biodefense Program for Pandemic Preparedness
OpenAI is opening GPT-Rosalind to vetted researchers and partnering with CEPI, Johns Hopkins APL, and Lawrence Livermore on pandemic preparedness.
OpenAI and Thrive Launch Self-Improving Tax AI Hitting 97% Accuracy
A Codex-powered tax system processed 7,000 returns at 30+ firms, rewriting itself from accountant corrections to climb from 25% to 86% accuracy in six weeks.
xAI Commits to Daily Grok Build Release Notes as Coding Agent Expands
xAI is shipping Grok Build updates daily as the coding agent leaves SuperGrok Heavy exclusivity to reach all SuperGrok and X Premium+ subscribers.
5WPR Study: AI's 15 Biggest Players Rank Worlds Apart on Transparency
First structured benchmark of how 15 leading AI companies communicate finds only one-third disclose user numbers, revenue, or enterprise customers.
Anthropic Opens Milan Office: Italy Becomes 6th European Market
Anthropic opens its Milan office on May 28, marking the sixth European city in under a year as EMEA revenue jumps 9x year-over-year.