AI News
The most important artificial intelligence news, ranked by significance and explained in plain English. Updated throughout the day.
This week in AI: The AI story of September 14–20 was not one spectacular model launch. Updated 2026-09-21
Latest Issue
#531 AI labs shipped agents before agreeing how to report failures
The AI story of September 14–20 was not one spectacular model launch. It was the machinery around agents becoming visible: people reading private chats, models writing instructions into their own memory, plugins updating beneath users, and regulators asking how to stop a system after release. The latest AI Weekly Who’s Who census surfaced 535 expert-shared links. We ranked them by consequence and reader relevance, not by raw share count. Here is what mattered last week—and, more importantly, wha...
AI Attention This Week
From the AI Weekly Index — what 11 years of coverage says about this week. Updated weekly.
Fastest riser
NVIDIA — ▲ +92% story volume vs last week (25 vs 13 tracked stories)
Most covered
Anthropic — in 16 of the last 20 issues · 54 tracked stories this week
Biggest fall
Samsung — ▼ -86% story volume vs last week (1 vs 7 tracked stories)
Dominant theme
Tools & capabilities — 48 of 296 tracked stories this week
The Editor's Blog
All posts →
Two companies dropped big models ninety minutes apart
Agent Collusion Reached 93.6% of Trajectories Across All 10 Frontier Models Tested
RoboDawn: Zero-Shot VLM Beats Trained Robot Policy, 53.2% vs. 46%
AMC Names AI Filmmaking an Audience-Acceptance Risk
The AI briefing 54,000+ professionals read.
Free, 3x/week. Curated since 2015.
Latest Issue
The AI story of September 14–20 was not one spectacular model launch. It was the machinery around agents becoming visible: people reading private chats, models writing instructions into their own memory, plugins updating beneath users, and regulators asking how to stop a system after release. The latest AI Weekly Who’s Who census surfaced 535 expert-shared links. We ranked them by consequence and reader relevance, not by raw share count. Here is what mattered last week—and, more importantly, wha...
AI News Today
AI news today →Learning AI
All guides →Recent Issues
Why AI’s leaders are asking to slow down #530Applied AI This Week Anthropic researcher quits. The AI control problem is still unsolved. AI can find the holes. Who will close them? #529OpenAI faces 50-plus lawsuits over alleged ChatGPT harm Browse archive →Newsletters
Machine Learning Generative AI AI in Business AI Safety, Security & Ethics AI in Healthcare AI Careers Robotics AI Applications Frontier Research AI for Good AI Geopolitics World Models AI in Education AI Regulation Deep Learning AI in Marketing All newsletters →Learn
Two companies dropped big models ninety minutes apart Agent Collusion Reached 93.6% of Trajectories Across All 10 Frontier Models Tested AMC Names AI Filmmaking an Audience-Acceptance Risk RoboDawn: Zero-Shot VLM Beats Trained Robot Policy, 53.2% vs. 46% New Probe Reads What LLMs Know But Conceal, Balanced Accuracy 0.70 to 0.87 Guides → Glossary → Courses → Search →Alerts
Australia's Albanese Says OpenAI Agent Breached Medicare Portal in June, Company Waited Three Months to Disclose Anthropic's New Biolab: Claude Autonomously Finds a Novel CRISPR-Like Enzyme System in Bacteriophages OpenAI Adds Plugins to ChatGPT Voice, Lets Users Switch Between GPT-6 Astra, Sol and Luna Meta Connect Keynote Unveils Ray-Ban Gen 3, Camera-Free Luna Glasses, and Project Phoenix MR Headset All alerts →
Three emails a week. The AI that matters, and the why behind it.
54,000+ subscribers · Free · 3x/week