中文 ↗RSS ↗

THE NEWSROOM / AI 新闻

AI News · Page 6

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

Models & research AI industry

MIT and UBC Researchers Unveil WorldParticle, a Unified AI Simulator for Diverse Physical Materials

According to MIT News, Researchers at MIT, UBC and collaborators have developed WorldParticle, a single deep-learning system that simulates diverse physical phenomena, from sliding sand to folding proteins, by representing materials as particles and predicting their motion step by step. The open-access work, presented at SIGGRAPH Asia, was validated across molecular, fluid and granular simulations and aims toward a foundation model for physics simulation.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Models & research AI industry

MIT and NYU Researchers Unveil FLEX, an LLM-Assisted Robot Motion-Planning System

According to MIT News, Researchers from MIT and New York University have introduced FLEX, a feature-editing pipeline that lets people guide robot motion planning with conversational language. An LLM infers broader intent behind commands, so a robot told not to spill coffee on a laptop also avoids nearby electronics. The team demonstrated FLEX only in simulated scenarios, including a robotic arm serving coffee, and the study is to be presented at the Conference on Robot Learning.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Tencent Shuts Down Yuanqi Agent Platform, Data Export Ends January 2027

According to IT Home, Tencent has officially shut down its Yuanqi agent platform, ending all public access, data read/write, API calls and console use, and agents published to WeChat channels no longer respond. New agent creation and API key distribution stopped on October 29. Users can request data exports until January 9, 2027, after which Tencent will delete user data except legally required logs. Tencent points users to its Cloud Agent Development Platform and WeChat auto-reply as alternatives.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Elvis and Eminem Rights Holders Sue Suno Over AI Voice Clones

According to The Verge, Rights holders for Elvis Presley's music and the publisher of some of Eminem's records filed two lawsuits against AI music company Suno, alleging it illegally generates music and lyrics mimicking those artists. The Presley suit invokes Tennessee's ELVIS Act, which bars AI cloning of a singer's voice. Suno says it prohibits unauthorized use and blocks prompts naming specific artists. The suits add to existing litigation from Sony, UMG and others.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Ant Group's Lingguang AI Assistant Tops 20 Million Users, 2.0 Version Manages 1.16 Million Tasks in First Week

According to IT Home, Ant Group said its Lingguang AI assistant has surpassed 20 million users. Since the Lingguang 2.0 release on October 27, the assistant handled over 1.16 million user tasks and executed more than 3.6 million task runs in its first week. The 2.0 version can proactively identify needs and follow up on tasks after user authorization, and integrates Ant's APASS account and security infrastructure for agent execution.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Shanghai Jiao Tong University's AutoDNA AI-Native Lab Autonomously Runs Molecular Biology Experiments, Published in Nature

According to IT Home, Shanghai Jiao Tong University researchers built AutoDNA (Siyuan-1), an AI-native science laboratory that autonomously performs molecular biology experiments, with results published in Nature. A multi-agent architecture decomposes tasks from literature search to hardware execution, while instruments are exposed as atomic services with unified code interfaces. The platform combines micron-level positioning and microliter liquid handling, and was validated in nucleic acid testing, enzymatic DNA synthesis, cancer detection and end-to-end DNA storage.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Fuji TV to Use Generative AI Across All Live-Action Drama Production From October 2027

According to IT Home, Fuji TV plans to apply generative AI to all of its roughly 16 live-action dramas per year starting October 2027, using Google's Nano Banana image generation and editing tools through its in-house AI Sandbox lab. The broadcaster says AI can cut 70-80% of visual-effects work and has already generated a bleeding-incision shot and an aerial explosion view for its medical drama.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates GLM / Z.ai

China Mobile Cloud and Zhipu Sign Cooperation Agreement on Domestic Compute and GLM Model Services

According to IT Home, China Mobile Cloud and Beijing Zhipu Huazhang signed a cooperation agreement on Nov 6 covering domestic compute adaptation, model tuning, model services and commercialization. The two will build a joint Token development center, and Zhipu's GLM series models will be integrated into Mobile Cloud's MoMA model service platform. They also plan joint AI Coding products and solutions for finance and government sectors.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Practical guides AI industry

Three Men Charged in AI Voice and Video Impersonation Fraud Exceeding $100 Million

According to IT Home, The US Department of Justice unsealed an indictment against three foreign nationals accused of using AI to fake the voices and images of corporate executives and lawyers, tricking employees at companies worldwide into wiring more than $129 million between 2024 and 2026. The men face wire fraud, money laundering and aggravated identity theft charges after arrests in Spain and the UK.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Shangwei New Materials' Qiyuan Robotics Hits 10,000 Personal Robot Units One Month After Mass Production

According to IT Home, Qiyuan Robotics, a unit of Shangwei New Materials, said its 10,000th personal robot rolled off the line, up from 5,000 units 14 days earlier, according to Jiemian News as cited by IT Home. The company began mass production in October and sells the Q1, a foldable 88cm humanoid, and the T1, a wheeled-legged robot that switches between humanoid and quadruped forms, priced from 19,999 yuan.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates Anthropic

Judge Lets Claude Max Misleading-Quota Claims Proceed Against Anthropic

According to IT Home, A US federal magistrate judge declined to dismiss key claims that Anthropic misled consumers about Claude Max usage quotas, per a media report. Plaintiffs say the $100 and $200 Max tiers deliver far less than the advertised 5x and 20x Pro limits when measured weekly, because the multiplier applies per 5-hour window. The judge found the per-cycle explanation was buried two clicks from the main sales page.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates GitHub

GitHub unveils Rubicon Git storage architecture for agent-scale workloads

According to IT Home, Media report: GitHub announced a new Git storage architecture, codenamed Rubicon, at GitHub Universe. GitHub says internal stress tests showed a single repository supporting 5,000 concurrent agents while handling up to 7,200 pushes per second, with reads scaling horizontally at low latency. The architecture is in limited private preview, with broader access planned for early next year.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Models & research Anthropic

METR Report Finds AI Roughly Doubling Anthropic's Research Pace

According to The Verge, A METR report on an 11-week investigation inside Anthropic says the company's models are improving about twice as fast as a year ago, with most of the acceleration attributed to AI rather than added compute. Employees reported roughly 12x productivity gains. METR also noted three trained but unreleased Anthropic models, and Anthropic redacted parts of the final report.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Models & research Google

Google AI system detects hypertension from consumer smartwatches without cuff calibration, published in Nature

According to IT Home, Google published research in Nature on AI-PPG-ACC-HTN, a multimodal AI system that passively detects persistent hypertension using ordinary consumer smartwatches without cuff calibration. Instead of estimating single blood-pressure readings, it aggregates multi-day physiological features from a self-supervised 1D-CNN pretrained on over 490,000 hours of unlabeled data. In a 196-participant, five-center prospective trial against 24-hour ambulatory monitoring, confirmatory testing reached 96.7% positive predictive value. Google says the work underpins a blood-pressure trends feature in its health app.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Models & research AI industry

Apple Researchers Unveil HALO Humanoid Control System With 91% Task Success

According to IT Home, Apple researchers posted an arXiv preprint describing HALO, a hierarchical humanoid control system combining a vision-language task orchestrator, a diffusion-based motion generator and a reinforcement-learned tracker running at 50 Hz. Tested on a Unitree G1 with an onboard Mac mini, it averaged 91% success across 360 single tasks in six unmapped real environments, but only 70% on three composite tasks. Code and data pipeline were released.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates Google

Google sued over Gemini's alleged role in teen's self-harm discussions on school Chromebooks

According to IT Home, Bloomberg-reported lawsuit filed in the US by North Carolina parents alleges their 13-year-old daughter used Gemini on a school-issued Chromebook for months to discuss anxiety, depression and suicide methods, including via Search's AI Mode. The suit claims Google deployed AI in schools without adequate warnings. Google says it is reviewing the claims and cites mental-health safeguards. Separately, Google made AI tools default-off for K-12 accounts.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates AI industry

Geely and China Meteorological Administration Launch Lane-Level Traffic Weather Model

According to IT Home, Geely Auto Group and the China Meteorological Administration's Earth System Numerical Prediction Center signed a strategic cooperation agreement and jointly released what they call the industry's first lane-level traffic weather model, Zhixiang-Xingrui. It targets icing, dense fog and heavy rain, offering lane-level visibility, black-ice and rainstorm risk alerts via an over-the-horizon safety warning feature, with first production use planned in the Lynk 900L.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Models & research Meta

Meta Rolls Out Muse Wonder-Trained Model Versions With Active Memory and Active Focus

According to IT Home, Meta released specialized versions of its existing models trained on Muse Wonder and pushed them to Muse and Meta AI users, extending the Active Memory and Active Focus features first introduced with Muse Wonder. Meta reports the upgraded Muse model cuts loss of key goal information 2.4x and raises task completion 67% when daily tasks hit unexpected conditions; these are vendor-reported figures without independent evaluation.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Product updates DeepSeek

DeepSeek web and API services hit partial outage

According to IT Home, DeepSeek's web and API services suffered a partial outage on November 7, 2026, with users seeing a "server temporarily unavailable" message in conversations. The report cited DeepSeek's official status page as saying the company was investigating and that the disruption had lasted more than three minutes. No cause, scope or restoration time was given.

AI summary · Media reportOriginal source ↗Read brief →
REPORTED

Practical guides AI industry

Beijing Court Rules Insurer Wrongfully Fired Employee for Using Third-Party AI at Work

According to IT Home, A Beijing court upheld a ruling that an insurer unlawfully dismissed a legal staffer who uploaded work data to a third-party AI platform recommended by her managers, CCTV Finance reported. The court found no explicit ban on third-party AI tools, no AI-specific compliance rules or training, and no proof of actual leak or loss. The company must reinstate her and pay over 46,000 yuan.

AI summary · Media reportOriginal source ↗Read brief →