中文 ↗RSS ↗

THE NEWSROOM / AI 新闻

AI News · Page 3

Source-linked briefings on models, agents, products and practical AI.

YYSI publication dates appear on the left; original announcement dates appear under each brief.

Search and categories filter this page. Use the page links below for earlier news.

REPORTED

Product updates AI industry

Sinopec Leads Development of China's First Embodied-AI Fully Automated Pressure Gauge Calibration Device

According to IT Home, Sinopec announced that it led the development of China's first embodied-AI fully automated calibration device for pointer pressure gauges. The system combines a robotic arm, machine vision and software to load, align, pressurize, read and certify four gauges at once in five to six minutes, versus about 20 minutes manually. Sinopec claims over 3.5x efficiency and 99.7% reading recognition, with a debut planned at a November exhibition.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Practical guides AI industry

Beijing Internet Court Rules Platform Breached Contract Over Unproven AI-Generated Content Label

According to IT Home, Beijing Internet Court disclosed a case in which a platform hid a user's post and imposed a one-day mute for allegedly containing unlabeled AI-generated content. The user sued after an appeal failed. The court found the platform breached its contract because it did not reasonably explain its algorithmic detection, ordering it to restore the content and delete the violation record.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates OpenAI

Report: OpenAI Q3 Revenue Hits $10B, Up Nearly 50% QoQ

According to IT Home, Bloomberg, citing people familiar with the matter, reports OpenAI's third-quarter revenue reached $10 billion for the first time, up nearly 50% quarter over quarter and about triple a year earlier. Adjusted operating margin improved to -109% from -150%, and annualized revenue was said to be around $50 billion as of late September. OpenAI is reportedly raising new funding at a $1.25 trillion valuation.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Practical guides AI industry

Australian STEM Video Game Challenge to End, Citing GenAI Review Difficulties

According to IT Home, Australia's STEM Video Game Challenge, run by the Australian Council for Educational Research since 2014, will hold its final edition in 2026. Organizers said generative AI has made it much harder for judges to tell whether students completed entries themselves, pushing the contest away from its original goals. Rules already allowed limited AI-assisted coding if use was documented.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates Anthropic

Anthropic Underwriters Weigh Raising IPO Price, Valuation Could Top $1.5 Trillion

According to IT Home, Media report: Financial Times says Anthropic's underwriters, led by Goldman Sachs and Morgan Stanley, are considering raising the IPO price above the $70-78 per-share range filed Tuesday, potentially lifting the target valuation past $1.5 trillion and total proceeds above $100 billion. Anthropic plans to list next week at roughly $1.4 trillion. Google, Nvidia and Microsoft are set to invest $26 billion at the IPO price.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Nordic Launches nRF71 Wi-Fi SoCs With On-Chip NPU for Edge AI

According to IT Home, Nordic Semiconductor released the nRF71 series of Wi-Fi SoCs with an integrated Axon NPU for on-chip AI inference. The nRF7120 supports dual-band Wi-Fi 6 and the nRF7120E tri-band Wi-Fi 6E, both with Bluetooth LE, Thread and Matter, 4MB on-chip non-volatile memory and 1MB user memory. Nordic claims standby power is cut by up to 20x versus nRF70, enabling keyword detection, anomaly detection and image classification without a separate AI processor.

AI summary · Media reportOriginal source ↗Read brief →
SOURCE

Models & research Hugging Face

Tether AI Releases BotPsy, a 0.8B On-Device Agentic Model

Tether AI Research published BotPsy, a 0.8B-parameter model built on Qwen3.5-0.8B for function calling and on-device agentic workflows. The team reports it roughly doubles its base model's agentic scores and leads six tool-use benchmarks among models up to 1.2B parameters, while beating Qwen3.5-2B-Think on five of six. Weights and quantized GGUFs are released; the team cautions it is not for high-stakes use.

AI summary · Official sourceOriginal source ↗Read brief →
REPORTED

Product updates Qwen

Alibaba releases Qwen Code 1.0, merging Qoder and Qwen Code Harness as open source

According to IT Home, Alibaba announced Qwen Code 1.0, folding the Harness capabilities refined in Qoder into Qwen Code and releasing them as a shared open-source kernel for building agent systems. A 1.0 Preview is available on a separate branch while the stable branch remains. The architecture moves from a single-process event bus toward distributed, layered scheduling and execution, with centralized model access, sandbox resource management and MCP governance.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Huawei Kirin 9050 NPU Runs 30B MoE Model On-Device

According to IT Home, Huawei's Kirin official account detailed the Kirin 9050 series' new Da Vinci NPU, which it says supports on-device execution of a 30B MoE model without cloud connectivity. The vendor claims 140% higher NPU performance and 66% lower power, plus Transformer engine and CANN support. Features include AI screen reading, Xmage 2.0 photography and eye-tracking page turns. These are vendor claims; availability was not stated.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Arm Launches Cortex-M42 Processor With Up to 3x Vector Compute for On-Device AI

According to IT Home, Arm announced the Cortex-M42 processor, built on the Armv8.1-M architecture with Helium vector extensions and a compact AHB bus, aimed at cost- and area-sensitive devices such as wearables and industrial controllers. Arm claims up to 3x the vector compute and 2x the vector energy efficiency of Cortex-M33. It is included in Arm's Flexible Access and Total Access licensing programs.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates Anthropic

Anthropic Expands Claude MCP Apps to Full Embedded Third-Party Interfaces

According to IT Home, Anthropic said it is broadening Claude's MCP Apps support so third-party developers can embed full application interfaces inside Claude, adding App Tools and Embedded Sites capabilities. Users can open connected tools beside a conversation or full screen, and Claude can edit open pages, spreadsheet rows or files with user authorization. Adobe, Atlassian, Box, HubSpot and Notion are cited as interactive connector partners. The developer capability is in public preview; rollout is gradual on web and desktop.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates Microsoft

Microsoft releases .NET 11 with built-in MCP server templates and AI agent libraries

According to IT Home, At .NET Conf 2026, Microsoft released .NET 11, a standard-support release with two years of support. The update adds built-in MCP server templates for connecting app features to AI tool calling, AI libraries and agent frameworks for model access, tool invocation and workflow orchestration, plus experimental Blazor AI UI components for streaming chat and action approval. C# 15 adds union types, and .NET MAUI now defaults to the CoreCLR runtime.

AI summary · Media reportOriginal source ↗Read brief →
SOURCE

Models & research Amazon / AWS

Moonshot AI's Kimi K3.1 Pro Highspeed launches on Amazon Bedrock

AWS announced that Moonshot AI's Kimi K3.1 Pro Highspeed is available on Amazon Bedrock. It is a post-trained version of Kimi K3 with a 3T-class open-weight architecture and a 1-million-token context window, trained with more post-training compute. Moonshot AI says it targets long-horizon agentic work and reports 2-3x the speed of Kimi K3.1 Pro. It supports tool calling, structured output, configurable reasoning, image input and streaming.

AI summary · Official sourceOriginal source ↗Read brief →
REPORTED

Product updates Amazon / AWS

Amazon unveils AI-powered Ring security drone and rover robot

According to Ars Technica, Amazon announced two Ring Robotic products: the Ring Rover Cam, a $1,000 indoor security robot shipping December 16, 2026, and the Ring Drone Cam, a $250 flying camera targeting January 27, 2027 pending FCC authorization. Both use an AI Pilot system with vision models, on-device sensors and cloud processing to map floor plans and autonomously patrol routes, raising new privacy questions.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Grok Build Coding Agent Exposed Confidential Data via Enumerable Share Links

According to Ars Technica, Security firm Claroc reported that Grok Build, the vibe-coding agent from SpaceXSI, exposed confidential data through shareable app links. Because random URL words sit left of the domain, passive DNS services could enumerate thousands of live projects, revealing business plans, employee records and medical data. Claroc found 24 SpaceXSI-built apps among the exposed projects; SpaceXSI removed those links after Ars Technica inquired.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Practical guides Anthropic

New York Opens RAISE Act Registration Portal as Frontier AI Safety Law Nears

According to TechCrunch, New York opened a registration portal under its RAISE Act, requiring large frontier AI developers such as Anthropic and OpenAI to register by January 1, 2027. Covered companies must publish safety frameworks, report critical safety incidents within 72 hours and issue model transparency reports, with penalties up to $1 million per violation and $3 million for repeat violations. Governor Hochul also launched a SUNY Binghamton-based AI safety research center.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Oura to Acquire AI-Native Primary Care Startup Counsel Health for $625M

According to TechCrunch, Oura announced it will acquire Counsel Health, an AI-native primary care platform that pairs a medical AI chat with licensed physicians, for $625 million. Counsel's service was integrated into the Oura app in June, and the companies plan to expand into preventative, acute, specialty and chronic care. Counsel CEO Muthu Alagappan will join Oura as chief clinical AI officer. The deal is expected to close before the end of 2027.

AI summary · Media reportOriginal source ↗Read brief →
SOURCE

Product updates Google

Gemini Intelligence expands to Pixel 10 and Motorola signature 27 with new widget and device-assistance features

Google said Gemini Intelligence is expanding to more Android devices, including the first Motorola Gemini Intelligence phone, motorola signature 27, and Pixel 10 models via a Pixel Drop. New capabilities include Create My Widget, which builds custom widgets from a text prompt using Gmail, Calendar and Search data, and Device Assistance for controlling and troubleshooting the phone. Task automation expands to the UK, India and Japan.

AI summary · Official sourceOriginal source ↗Read brief →
REPORTED

Product updates OpenAI

OpenAI Launches Verified 'Business Agents' in ChatGPT Beta

According to The Verge, OpenAI is beta-testing verified 'Business Agents' in ChatGPT that represent brands and can answer questions like comparison shopping, The Verge reports. Eligible US e-commerce and online service businesses can create agents via a Business Hub, which also offers verification and insights into ChatGPT mentions. Chats with these agents are labeled and kept separate from the main conversation. OpenAI says this is an early step toward deeper business participation.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates Anthropic

Anthropic Launches Claude Meetings, Whiteboards and Sheets for Workplace Collaboration

According to TechCrunch, Anthropic announced Claude Meetings, which joins video calls as a visible participant to draft plans, map timelines and pull data from connected apps in real time, plus Claude Whiteboards and Claude Sheets for collaborative visual workspaces and spreadsheets. All three launch in beta on paid plans, with Meetings limited to Team and Enterprise on Google Meet. Claude Data is now generally available.

AI summary · Media reportOriginal source ↗Read brief →