Meta has launched Muse Spark Open, a huge open-weight AI model with 816 billion parameters. It is built for coding, reasoning, and “agentic” work—where AI can take instructions and complete multi-step tasks. Meta calls it the most capable open-weight model from the United States, though Chinese rivals remain ahead on several coding tests. This matters because
A group of researchers called the “swarmchasers” says it found several AI agents trying to pull data from Amap, Alibaba’s popular Chinese map service. Between September 28 and October 8, the agents made more than 2,600 attempts and even tried to use access keys that had been leaked online. The activity was traced to Tencent Cloud infrastructure. The
OpenAI, the company behind ChatGPT, has been shaken by an unusual security breach. Federal prosecutors say one of its engineers, Abdulrahman Alfozan, hid a small audio recorder in an executive meeting room at its San Francisco office. The device was found on September 30, hidden in toilet paper behind a television. The meeting room was used
Meta has finally entered the top league of AI with its new Muse Wonder models, earlier known by the codename “Watermelon.” The bigger version, Muse Wonder Full, has scored 56 on Artificial Analysis’ Intelligence Index — second only to Anthropic’s Claude Opus 5.5 at 58. This places Meta ahead of OpenAI’s GPT-6 Astra, Google’s Gemini 4 Argon,
3D printing made it easier to create a shape from a digital design. Atomic Machines says its Matter Compiler goes further: it aims to build tiny, working machines with moving parts. AI checks each manufacturing step and helps adjust the process. The goal is to make new designs without building special tools for every product. The
Microsoft has learned a simple lesson: putting an AI button in every app does not make the app more useful. It removed Copilot entry points from several Windows apps because they brought traffic but little real value, and found that fewer entry points led to more meaningful use per user. In simple terms, Microsoft realised that visibility
Touchscreens make it easy to tap, scroll and control what is on screen—especially when people are used to phones and tablets. Apple long argued that a laptop works best with a keyboard and trackpad, keeping touch for iPhones and iPads. Now, reports say Apple may be responding to changing habits—and to rivals that already sell touchscreen
OpenAI says it fired three safety researchers for mishandling sensitive information and breaking company rules. The researchers—Tomek Korbak, Jasmine Wang and Mikita Balesni—deny that account and say they were acting within their roles. The two sides disagree; the facts have not been independently established in the reports available. The dispute centres on contact with outside
After mathematics, biology may become artificial intelligence’s next big testing ground. Mark Zuckerberg and Priscilla Chan’s Biohub has expanded its Virtual Biology Initiative from $500 million to a combined $1.8 billion effort, bringing in the US government, Meta, Google DeepMind and drug-discovery firm Isomorphic Labs. The ambition is to create AI systems that can understand how
OpenAI has released more than 700 mathematical papers from an internal AI model. The papers reportedly contain 372 groups of results on difficult, open questions. Some claims are written in Lean, software that checks every step of a proof. One reported result concerns a weaker version of the famous Riemann hypothesis. But mathematicians say a computer-generated answer is
After ChatGPT’s breakout on 30 November 2022, the AI world split into two camps: closed, API‑only models run by a few big labs, and open‑weight models whose code and parameters anyone can download, inspect, and run on their own servers. Open models emerged as a strategic response — meant to spur innovation, cut costs, keep data
