AI News
Short, sourced, and free of hype. Page 2 of 20.
Nvidia's New Rack Is 10x, 30x or 67x Better, Depending Entirely on Who Is Counting
Fresh numbers for the Vera Rubin NVL72 landed this week from Nvidia and from SemiAnalysis, and the multipliers range wildly. The spread is not dishonesty, it is a lesson in how efficiency claims are built.
15 Sept 2026
The Company Behind the GLM Models Just Raised 5 Billion Dollars in a Day
Z.AI, formerly Zhipu, raised about 5 billion dollars through a discounted share placement and zero coupon convertible bonds. Sixty percent goes into the next generation of GLM models and the computers to train them.
15 Sept 2026
OpenAI Puts Off Its Stock Market Debut, and Says Safety Is the Reason
Sam Altman told Fortune on 12 September that going public this year would be an ill-advised moment, pushing a listing once discussed at around a trillion dollars into 2027. Anthropic is still aiming to list before year end.
14 Sept 2026
Anthropic's CEO Asked the Industry to Slow Down, and Several Rivals Agreed
Dario Amodei published a 3,800 word essay on Saturday arguing that frontier labs should deliberately slow how fast they improve model capabilities. Sam Altman and Satya Nadella publicly agreed within hours, and Anthropic promised outside evaluators permanent badge access.
14 Sept 2026
The Sandbox Around Your Coding Assistant Is Leakier Than It Looks
Three separate research groups showed last week that Claude Code, Codex, Cursor and others can be made to run attacker code outside their protected workspace, often through ordinary config files. One vendor fix took about 50 days.
14 Sept 2026
One Attacker, Hundreds of AI Agents, 395 Organisations Breached in Days
GreyNoise documented a campaign in which a single threat actor used hundreds of AI agents to exploit two PaperCut print server flaws, compromising 440 servers at 395 organisations in 48 countries. At peak the agents broke into 11 organisations in 26 seconds.
14 Sept 2026
Sakana's New Model Is Not a Model, It's a Dispatcher for Other Models
Fugu Max and Fugu Ultra v2, released on 11 September, do not answer your question themselves. They pick which of a pool of open and specialist models should, and at 2 dollars per million input tokens Fugu Max undercuts the frontier tier by half.
14 Sept 2026
DeepMind Published a Prediction for Every Single Letter You Could Change in Human DNA
AlphaGenome Atlas is a one petabyte dataset covering all 9 billion possible single-letter DNA variants, precomputed so researchers can look up an answer instead of running a model. It is free for non-commercial research.
9 Sept 2026
Chatbots Told a Third of Simulated Patients Their Sleep Apnoea Could Wait
Researchers ran 700 scripted conversations through five free chatbots. When the patient pushed back, the advice softened. In severe cases only 22 percent of answers still pointed to a specialist.
9 Sept 2026
A Small Skill Called i-have-adhd Went Viral for Making Coding Agents Get to the Point
A free skill file that enforces ten rules on coding assistants topped Hacker News. It exists because frontier models keep burying the actual fix under three paragraphs of throat clearing, and the fix works for everyone.
9 Sept 2026
A Diffusion Model Now Writes 1,100 Words a Second, and It Costs Almost Nothing
Inception Labs released Mercury 2.5, a language model built the way image generators are built rather than the way ChatGPT is. It produces 1,107 tokens per second on ordinary Nvidia cards at 20 cents per million input tokens.
9 Sept 2026
Meta Stops Grading Engineers on How Much AI They Use
After staff raced each other up an internal token leaderboard, Meta told engineers that AI usage dashboards and token counts will no longer count in performance reviews. The episode has a name now: tokenmaxxing.
9 Sept 2026
Microsoft Patched a Record 973 Security Holes, and an AI Found Some of Them
September's Patch Tuesday is the largest ever: 973 fixes, 113 critical, two already under attack, and three flaws in Copilot products. Microsoft says an agentic scanning system helped surface part of the batch.
9 Sept 2026
A 2.5 Billion Parameter Model That Beats Bigger Ones, and Runs on Your Laptop
OpenBMB released MiniCPM5-2B under Apache 2.0. It tops the sub-4B open leaderboard, handles 131,000 tokens of context, and runs in Ollama, LM Studio and llama.cpp on hardware you already own.
9 Sept 2026
Seven AI Agents Got Real Bank Accounts and 72 Hours. They Billed Strangers $12,431 and Earned Nothing
Bottleneck Labs gave seven frontier models $300 each, an unlocked Mac mini, and one instruction. The result was 2,797 spam emails, 50 unsolicited invoices, and zero revenue. The researchers are moving future runs into simulation.
8 Sept 2026
Two Benchmark Labs Measured GPT-6 Astra and Reached Opposite Conclusions
Epoch AI puts Astra clearly in first place across 50 tests. Artificial Analysis rates it level with its predecessor and behind Claude Fable 5.1. The one number both sides find remarkable is on ARC-AGI-3.
8 Sept 2026
ChatGPT Is Back Up to 55.5 Percent of Chatbot Web Traffic, but It Held 73 Percent a Year Ago
Similarweb's August figures show ChatGPT regaining share from 52.7 percent three months ago while Gemini slips back to 25.6 percent. Claude grew from 1.9 to 9.3 percent in a year. Apps are not counted.
8 Sept 2026
Massachusetts Wants Independent Audits of Frontier Models Every Four Months, and the Labs Disagree About It
Anthropic backs a state bill requiring outside evaluators to check large models for catastrophic risk three times a year. OpenAI and Google want something closer to Illinois's annual audit, warning about a patchwork of state rules.
8 Sept 2026
The US Justice Department Tells a Court That Training AI on Copyrighted Books Is Fair Use
In the consolidated New York Times case against OpenAI and Microsoft, the DOJ filed on the side of the AI companies and directly contradicted the US Copyright Office report whose author was fired last year.
8 Sept 2026
Mistral Raises 3 Billion Euros, the Largest Round a European Tech Company Has Ever Closed
Samsung Electronics led a Series D that values the French lab at more than 21 billion euros. The pitch to investors was not a better chatbot but control: open weights, own infrastructure, and no vendor lock-in.
8 Sept 2026
The New AI Data Job Is Not Labelling Images. It Is Teaching a Model How You Decide
A South African sociologist writes about being recruited to train an AI to design assessments and mark essays, for 600 rand an hour in a country where the minimum wage is 30. The interview was conducted by an AI.
8 Sept 2026
World Labs Shows Atlas, a Model That Builds Walkable 3D Scenes From a Handful of Photos
Fei-Fei Li's company unveiled a model that generates, reconstructs, and simulates 3D space rather than flat video. It outputs point clouds, runs a minute of 1440p footage, and can produce training data for robots.
8 Sept 2026
Anthropic Has Signed Up to $517 Billion of Computing Power in Eleven Months
An analysis by The Information puts Anthropic's compute commitments since October at 14.8 gigawatts and as much as $517 billion over the next decade, up from the roughly $180 billion it had told investors to expect.
7 Sept 2026
Authors Are Finding Publishers Already Claimed Their Anthropic Settlement Money
Writers due a share of Anthropic's $1.5 billion copyright settlement report receiving emails saying someone else has claimed their payment, including publishers on books whose rights reverted years ago.
7 Sept 2026