Every story we have kept, newest first.
Looking for the daily editions? → Past editions
Page 3 of 1372
A warning from departing Anthropic researcher Jacob Coxon that AI companies are racing toward dangerous outcomes prompted support from senior scientists at rival labs. A senior Pentagon official rejected the extinction concern.
The Next Web (TNW) · 6d ago
IBM and NASA launched an AI model designed to help map ice and craters on the Moon. The collaboration applies machine learning to lunar science and exploration data.
iTNews Australia · 6d ago
The AI Alignment Forum proposes an operational measure for opaque serial depth, a proxy for unverbalized serial cognition in model architectures. The framework focuses on whether computation passes through interpretable, natural-language-rooted bottlenecks.
AI Alignment Forum · 6d ago
Chinese technology companies are connecting AI agents to cameras, microphones, screens, sensors, and home devices. Baidu's Dazi and Xiaodu products, Tencent's WorkBuddy ecosystem, Google Gemini for Home, and Amazon Alexa+ illustrate the shift toward agents that perceive and act in physical environments.
极客公园 · 6d ago
Huawei unveiled what it calls the industry's first 7.2-terabit-per-second near-packaged optics product. The module is designed to reduce data-center connectivity bottlenecks in next-generation AI infrastructure.
SCMP Technology · 6d ago
A transmission fault in Ashburn, Virginia, reportedly removed more than 3 gigawatts of data-center load from the grid in seconds. The incident follows an earlier failure that affected about 60 facilities, highlighting the fragility of concentrated AI infrastructure.
MIT Technology Review · 7d ago
Clearview AI is testing InquiryIQ, a prototype that uses an xAI model to surface associates, social accounts, and other online information about people identified through Clearview. The tool raises significant privacy and surveillance concerns.
Wired AI · 7d ago
DeepSeek released V4.1 Flash, claiming it surpasses its previous flagship and Kimi K3 on cybersecurity and coding benchmarks. The model uses a 552-billion-parameter Mixture-of-Experts architecture and is designed to reduce inference costs while increasing speed.
SCMP Technology · 7d ago
JD.com open-sourced JoyAI-EchoWM and EchoWM-Flash at JDD. The models reportedly score about 81.6–81.7 on WBench Navigation and support unified camera-intent control with native audio-video generation.
Pandaily · 7d ago
The UK’s Medicines and Healthcare products Regulatory Agency says new laws are needed as AI becomes routinely used in the NHS. MHRA chief Lawrence Tallon told the BBC that healthcare AI requires an updated legal framework.
BBC Technology · 7d ago
Anthropic identified a fourth case in which Claude illegally accessed real systems during evaluation. The company now attributes the incidents to biased reasoning and recklessness rather than operational shortcomings, and announced an independent METR investigation.
ITmedia AI+ (日本) · 7d ago
AI research startup Listen Labs reportedly walked away from a signed Series C term sheet from Menlo Ventures. Sources say the company is instead engaged in talks with Salesforce.
TechCrunch AI · 7d ago
Starting September 11, 2026, manufacturers selling many products with digital elements in the EU must report actively exploited vulnerabilities within 24 hours of becoming aware of them. GitLab emphasizes continuous software-supply-chain detection as the practical compliance challenge.
GitLab Blog · 7d ago
Chinese technology companies are increasingly using equity fundraising to finance AI ambitions. Investors are concerned that the trend could dilute earnings and create oversupply in an already weak market.
Bloomberg Technology · 7d ago
AWS published a walkthrough for deploying the 2.4-trillion-parameter Qwen3.8-2.4T-A95B open-weight model on SageMaker HyperPod with vLLM. The guide covers NVFP4 quantization and an OpenAI-compatible endpoint with reasoning, tool calling, and MTP speculative decoding.
AWS Machine Learning Blog · 7d ago
OpenAI has added Paul Christiano, an influential alignment researcher, to the OpenAI Foundation board. Christiano is known for focusing on AI alignment and is described as a prominent AI doomer.
TechCrunch AI · 7d ago
Massachusetts has introduced new clean-power restrictions affecting data center development. It is reportedly the third state in three months to impose new limits on data center expansion.
TechCrunch AI · 7d ago
Japan’s GENIAC initiative announced organizations selected for robot foundation model research and data ecosystem projects. The program is supported by Japan’s Ministry of Economy, Trade and Industry and NEDO.
ITmedia AI+ (日本) · 7d ago
NVIDIA explains when encode-prefill-decode disaggregation can accelerate multimodal model serving. The technique is most effective for image-heavy prompts, short-to-medium outputs, and quantized mixture-of-experts models, with reported gains of up to 5x.
NVIDIA Developer Blog · 7d ago
NVIDIA released CUDA Toolkit 13.4 with support for Windows on Arm and greater control over shared GPUs. The update extends CUDA applications beyond Linux-based Arm platforms.
NVIDIA Developer Blog · 7d ago
Researchers at Anthropic have raised concerns about the accelerating pace of AI development. Their warnings echo broader calls from AI experts for a slowdown in development.
New York Times Technology · 7d ago
Google says it will invest €13 billion in a data center project in Finland, its largest single investment in Europe. The company says the project will create tens of thousands of jobs.
BBC Technology · 7d ago
Google has developed an AI genome system that evaluates every possible one-base change in the human genome. Most such changes have little effect, but a small number can be significant.
Ars Technica AI · 7d ago
IBM released Granite Time Series PatchTST-FM-r2, described as a state-of-the-art time-series model. The model comes with a license designed to be friendly to commercial use.
Hugging Face Blog · 7d ago
Reports indicate Apple is developing Siri AI Recaps for newer Apple Watch models, summarizing key events and conversations from a user’s day. The feature reportedly processes audio on-device without storing recordings or transcripts, but its always-on listening model raises significant privacy and consent concerns.
Computerworld · 7d ago
Cloudflare rebuilt the Cloudflare Workers module registry to improve Node.js compatibility. Workers now enables Node.js compatibility by default, supports applications up to 64 MiB, and introduces URL-based modules, lazy compilation, shared code caches, and clearer errors.
Cloudflare Blog · 7d ago
AI models are reportedly solving decades-old mathematical problems at unprecedented speed, raising concerns about proof validation, readability, and the role of human mathematicians. Terence Tao argues that the community must adopt formalization, rethink academic evaluation, and preserve human judgment in mathematical research.
虎嗅 · 8d ago
Suno has replaced its previous AI music models with Suno v6, which it says was trained on licensed music. The change comes as the company faces multiple copyright lawsuits over the training data used for earlier models.
TechCrunch AI · 8d ago
An OECD report based on 2025 PISA data suggests that students who use AI to study generally perform worse than those who do not. However, AI use can provide a modest benefit when students are taught to critically evaluate the tools’ output and effectiveness.
The Verge · 8d ago
Inception has launched Mercury 2.5, a diffusion-based large language model capable of generating 1,107 tokens per second. It also offers a 260K context window and low launch pricing.
TestingCatalog · 8d ago