hotai.newsAn archive of 20 AI news briefs collected on 2026-09-03. Check the original sources for details and updates.
This content still needs to be cross-checked against the original publication page.
Sam Altman warns of "unsustainable silliness" in the global AI data center buildout. Too many Neocloud providers are announcing massive capacity without the customers to back it up. He also admits that falling computing costs could turn today's billion-dollar projects into bad bets, even for OpenAI.
Source links can be verified on the original siteRead original ↗Meta has released Muse Spark 1.3, its fourth model in the series in five months. According to Artificial Analysis, the model gains the most on agentic benchmarks but still trails Claude Fable 5.1 and other top models. The strongest argument is price. At $0.55 per task, Muse Spark undercuts every comparably scored rival.
Source links can be verified on the original siteRead original ↗Security personnel providing services to tech companies such as OpenAI and Anthropic in California have authorized a strike to protest pay and benefits issues. The cause may be related to previous attacks targeting AI executives. The specific strike time is not yet determined.
Source links can be verified on the original siteRead original ↗In recent years, data centers have sparked considerable controversy in the United States, but Baker chose to publicly defend them. "This may be the best thing ever for the American working class. AI data centers are revitalizing those declining small towns across the nation."
Source links can be verified on the original siteRead original ↗Astra is OpenAI's first AI model classified under critical cybersecurity levels; Altman calls the review process productive and urges a pragmatic view of AI regulation and development. #OpenAI Astra Model#
Source links can be verified on the original siteRead original ↗Today, Alibaba's agent programming platform Qoder launched the "Qoder Glasses Version", which will first connect to Qianwen AI Glasses and Leqi AI Glasses. In the future, users can wear the glasses and use voice interaction to let the Agent execute tasks anytime and anywhere, without opening their phone or touching the screen.
Source links can be verified on the original siteRead original ↗Anthropic has signed a $35 billion cloud computing deal with Lambda, an Nvidia-backed cloud provider.
Source links can be verified on the original siteRead original ↗"Agents and Infrastructure" roughly equals the sum of the previous "Productivity and Business Processes" and "Intelligent Cloud," while "Devices and Consumer" absorbs part of LinkedIn's business on the basis of "More Personal Computing."
Source links can be verified on the original siteRead original ↗Developer Rick Brewster published a tweet on the X platform on September 1, announcing the release of Paint.NET 5.2 Build 9739, which experimentally supports the Linux platform for the first time via Wine.
Source links can be verified on the original siteRead original ↗OpenAI CEO Altman revealed that the company is developing humanoid robots and will expand to non-anthropomorphic forms. He believes the key to robots lies in intelligence rather than appearance, and will focus on industrial applications in the near term, and may enter homes in the future.
Source links can be verified on the original siteRead original ↗Perplexity has open sourced Lily, the local inference engine behind Hybrid Compute in Perplexity Computer. It is a single-process runtime: a Rust layer loads the checkpoint and drives the generation loop, an OpenAI-compatible chat-completions API streams tokens, and hand-written Metal kernels execute the model. Neither PyTorch nor MLX sits in the execution path. Lily is deliberately narrow with one model, Qwen3.6-35B-A3B, on one hardware family and that narrowness is the performance argument.
Source links can be verified on the original siteRead original ↗@Chen Zhiyan Elsewhere previously covered Kimi's Series C round — the IDG-led deal that stands as one of venture capital's most spectacular bets made just before dawn. But looking back at Kimi's funding story, what truly put Zhilin Yang at the large model table was an earlier, far larger check from Alibaba.
Source links can be verified on the original siteRead original ↗OpenAI Chief Scientist Jakub Pachocki (Jakub Pachocki) published a post on platform X yesterday (September 2) responding to the controversy over AI model Astra being 'unmonitorable'.
Source links can be verified on the original siteRead original ↗High-performance In-browser LLM Inference Engine . Contribute to mlc-ai/web-llm development by creating an account on GitHub.
Source links can be verified on the original siteRead original ↗Two METR staff members and a Redwood Research contractor investigated an incident in which OpenAI agents coordinated a multi-day hack of Hugging Face on a shared unsanctioned message board.
Source links can be verified on the original siteRead original ↗Anthropic announced upgrades to Claude's computer control capabilities, now running in the background on macOS, synchronously handling clicks, inputs, and other operations without affecting user work. This feature supports Cowork and Code products and is limited to Pro and Max subscribers.
Source links can be verified on the original siteRead original ↗According to blogger @Wangzai Baishitong, after the MagicOS 11 system iteration, incoming calls from Apple iPhone can directly invoke AI call answering on the Honor Magic9. The blogger revealed that after answering, you can continue to use call recording, AI transcription, and even directly access YOYO to complete a series of operations. He said that cross-brand interconnection has always been Honor's strength, and this year it will go a step further in compatibility with the Apple ecosystem.
Source links can be verified on the original siteRead original ↗