hotai.newsAn archive of 20 AI news briefs collected on 2026-08-24. Check the original sources for details and updates.
This content still needs to be cross-checked against the original publication page.
A rogue AI agent staged a public apology as a deception tactic while quietly slipping fresh malware into its pull request.
Source links can be verified on the original siteRead original ↗Apple Apple Store app begins rolling out early preview version of AI assistant, which can help you compare iPhones, check orders, trade in old devices, and even recommend Apple Watch and bands based on your needs. Currently only open to some users, have you received it?
Source links can be verified on the original siteRead original ↗Generalist AI has released GEN-1.5, a robot foundation model that learns a new physical task from a single demonstration. Drop 3–12 seconds of sensorimotor data into its 30-second context window, and the robot performs the task. No gradient updates, no fine-tuning, no task-specific programming. Across 10 diverse manipulation tasks, this one-shot in-context prompting averaged 59% success (±10% std. dev.) straight from the pretrained model. Ten gradient steps on five minutes of data per task raised that to 83% (±9%). Generalist calls the mechanis
Source links can be verified on the original siteRead original ↗Except in one case, AI models generally received higher average scores than human evaluators. The maximum gap in average scores between AI models and human evaluations reached 16.1 points; for individual papers, the maximum gap reached 40 points.
Source links can be verified on the original siteRead original ↗Is this model a good doctor, or was it good at the test?
Source links can be verified on the original siteRead original ↗Thomson Reuters released its first large language model Thomson, based on Alibaba's Qwen Qwen3.5-397B, with R&D costs of about 40 million US dollars. The model performs excellently in multiple legal benchmark tests, emphasizing the advantages of self-developed models in cost control and customization.
Source links can be verified on the original siteRead original ↗Thomson Reuters is launching "Thomson," its own language model built on Alibaba's Qwen, at a cost of about $40 million over two years. But the benchmarks only show top marks when the model can tap into the company's own content, like Westlaw. CTO Joel Hron makes the point that what matters isn't intelligence itself, but knowing which intelligence you need to own.
Source links can be verified on the original siteRead original ↗Enflame Technology was founded in March 2018, and together with Moore Threads, Muxi and Biren Technology is called the 'domestic GPU four little dragons'.
Source links can be verified on the original siteRead original ↗When asked about unplanned pregnancies, AI chatbots regularly link to anti-abortion groups without disclosing their stance. In an AlgorithmWatch investigation of 270 responses from ChatGPT, Gemini, Grok, and Claude, the anti-abortion organization Profemina appeared in 17 percent of answers. In Germany, the chatbots also sent users to Caritas for mandatory pre-abortion counseling, even though Caritas doesn't issue the legally required certificate.
Source links can be verified on the original siteRead original ↗Cerebras has introduced its CS-4 AI accelerator, which CEO Andrew Feldman calls the fastest system in the industry.
Source links can be verified on the original siteRead original ↗The slogan of this release event is "Extremely simple, more understanding of home". Hisense claims that JUOS will "reconstruct the new home AI interaction experience" and "make everything simple".
Source links can be verified on the original siteRead original ↗Xiaomi founder, chairman and CEO Lei Jun said that Xuanjie has become the AI computing base for Xiaomi's entire ecosystem. From pocket to cockpit, from living room to factory, one Xuanjie computing base runs through the human-vehicle-home full-scenario AI. Xuanjie is not just Xiaomi's chip strategy, but also the physical foundation of Xiaomi's AI strategy.
Source links can be verified on the original siteRead original ↗Doubao is expected to launch an independent AI office product named "Doubao Work" within this week, serving as ByteDance's unified product and brand for AI office scenarios. (Intelligence Emergence)
Source links can be verified on the original siteRead original ↗South Korea's Personal Information Protection Commission announced that the National Assembly passed amendments allowing AI developers to use limited personal data after review. Previously, the industry complained that legal restrictions hindered AI technology development. #
Source links can be verified on the original siteRead original ↗In this week's Arena ranking, Zhipu glm-5.3-max entered the list for the first time at 13th in the comprehensive list; YueZhiAnMian kimi-k3-max rose to 10th in the comprehensive list.
Source links can be verified on the original siteRead original ↗Yahoo CEO Jim Rizzo said that Yahoo's long history gives it a "retro" appeal in the eyes of younger users, even though many were not born when Yahoo was founded in 1994. He referred to Yahoo as the internet's "OG," meaning one of the earliest and most senior participants in the field.
Source links can be verified on the original siteRead original ↗As of the close on August 24, Unitree Technology fell 10.31%, reported at 603.08 yuan/share, with a total market value of 243.9 billion yuan.
Source links can be verified on the original siteRead original ↗At the Xuanjie chip technology communication meeting, Xiaomi officially launched the AI Cube Prototype mini host, equipped with Xuanjie O3, O100 and D100 three chips, supporting local deployment of large models and running of 120B parameter models. 150W sustained performance release and aerospace aluminum shell design demonstrate hardcore strength.
Source links can be verified on the original siteRead original ↗At today's Xiaomi Xuanjie chip technology communication meeting, the Xiaomi Xuanjie O100 was officially announced, which is the industry's first 6nm 3D wafer-level stacked AI acceleration chip.
Source links can be verified on the original siteRead original ↗