AI news, Nov 11, 2026: Gemini 4 Carbon Launch, OpenAI's $1B Quarter, Anthropic IPO Push
Google DeepMind has officially released the Gemini 4 Carbon model, enhancing performance in complex programming and multi-agent scenarios. OpenAI's third-quarter revenue exceeded $10 billion for the first time, though adjusted losses expanded to $10.9 billion. Meanwhile, Anthropic launched an AI office suite and is preparing for an IPO, with reports suggesting a 20% share allocation for retail investors, signaling intense market attention.
Lead written by AI from the day’s stories
Top stories
- Models
Gemini 4 Carbon Now Available
Google DeepMind announced the official release of the Gemini 4 Carbon model. This update to the Gemini series further enhances performance and availability in specific contexts. The launch marks Google's continued iteration in generative AI models.
The VergeGoogle DeepMindHacker News9 outlets
- Models
DeepMind Launches AlphaGenome 2 to Analyze DNA and Disease
Google DeepMind released AlphaGenome 2. The model predicts how DNA sequences and variants impact specific human cell types and diseases. This advancement aids in understanding disease mechanisms.
The VergeGoogle DeepMindIT之家4 outlets
- Products
Anthropic Launches AI Office Suite
Anthropic has launched an AI-powered office suite, expanding its portfolio of enterprise tools. This move allows the company to broaden its reach beyond coding and high-end corporate clients.
The VergeTechCrunchTHE DECODER4 outlets
- Research
Google DeepMind mathematically verifies key parts of Signal
Google DeepMind completed a four-week project to mathematically verify key parts of the Signal protocol. The result suggests that AI can make formal verification significantly more practical.
Ars TechnicaiThomeTHE DECODER3 outlets
- Funding
OpenAI Quarterly Revenue Hits $10B, Trails Anthropic
OpenAI grew quarterly revenue 50% to $10 billion, but adjusted operating losses rose to $10.9 billion. It trails rival Anthropic in revenue scale and is engaging in enterprise price wars to close the gap.
THE DECODERTechCrunch2 outlets
Models15
- Models
DeepL introduces Aure model for emotional real-time translation
DeepL has introduced its new Aure model, designed for real-time voice translation. The model preserves the tone, emotion, emphasis, and speaking pace of the original speaker in the target language, aiming to handle complex pragmatics such as sarcasm more naturally than traditional tools.
THE DECODERIT之家2 outlets
- Models
Alibaba Qwen4 Preview Launches, Focusing on Coding
Alibaba announced the upcoming Qwen4 model, with a preview version now available. The new model enhances knowledge understanding, world perception, and coding capabilities, and is integrated into multiple applications.
IT之家ITmedia AI+2 outlets
- Models
Google launches Gemini 4 Carbon
Google announced the general availability of the Gemini 4 Carbon model on November 10, positioning it as a high-performance AI for complex coding and research. The model supports multi-agent collaboration features and is integrated into paid plans.
- Models
Meta Unveils Muse Ember Lightweight Model
Meta released Muse Ember on Nov 10, a lightweight AI model designed for autonomous agents. It offers superior performance cost efficiency, matching Claude Haiku 5.5 in benchmarks, aiming to lower deployment barriers for enterprises.
- Models
Google Releases Gemini Robotics 2.5
Google DeepMind released the robotics AI model Gemini Robotics 2.5 on November 10, 2026. The model is designed to accelerate learning in the real world and adapt to various scenarios, including daily life and factory settings, enhancing robotic adaptability.
- Models
Nvidia GR00T 2 Learns Robot Control from Human Video
Nvidia released its robotic AI model GR00T 2, which trains robotic arms on nearly 400,000 hours of human video to perform tasks without manual teleoperation. The system's success rate currently ranges from 10% to 80%, indicating it is still developing.
- Models
Ernie New Model Creates Backrooms in Pure Code, Switching 20 Art Styles in 116 Seconds
Baidu's Ernie has released a new model capable of creating Backrooms content in pure code and switching between 20 art styles in 116 seconds. The model will be open-sourced, demonstrating strong code generation and creative capabilities.
- Models
NaiveAI claims 2,100 tokens/sec speed; coding tested
NaiveAI claims a throughput of 2,100 tokens per second, potentially topping speed rankings. Hands-on tests suggest its coding quality is comparable to DeepSeek, making it attractive to speed-focused users.
- Models
Fish Audio Releases Live 3 Full-Duplex Voice Model
Fish Audio released Live 3, a full-duplex voice model designed for voice agents. It listens while speaking, distinguishes between backchanneling like 'mm-hmm' and actual interruptions, and reacts in as little as 186 ms. The model currently supports English only.
- Models
Google Releases Gemini 4 Carbon Model to Compete on Price
Google announced that its new frontier model Gemini 4 Carbon is now fully available to enterprises and developers. The model's API cost is half that of Anthropic's Fable/Mythos 5.2, focusing on enhanced capabilities in coding, agentic AI, and long-duration autonomous task execution.
- Models
Datadog releases dedicated AI model for SRE troubleshooting
Datadog has released a dedicated AI model tailored for SRE troubleshooting scenarios. The model is designed to help operations personnel diagnose system issues more efficiently, improving infrastructure monitoring and maintenance.
- Models
Google Releases Gemini 4 Carbon, Enhancing Agents and Security
Google officially released its latest AI model, Gemini 4 Carbon, for developers, enterprises, and consumers. The model focuses on enhancing cybersecurity, complex software engineering, scientific research, and long-running AI agent workflows to improve performance on high-difficulty tasks.
- Models
Google Releases Gemini Robotics 2.5 for Physical AI and Robotics
Google released its latest VLA robot model, Gemini Robotics 2.5, enhancing dexterity and reliability in unstructured environments. The model has been tested on a Mercedes-Benz assembly line, helping Apptronik robots achieve a 96% success rate in picking tasks.
- Models
Anthropic Launches Claude Fable 5.2 and Mythos 5.2
Anthropic released flagship model Claude Fable 5.2 and specialized model Mythos 5.2, featuring a 30% speed boost and a 20% API price cut. The models target complex open-ended tasks, outperforming predecessors in code refactoring while optimizing token consumption to reduce deployment costs.
- Models
MiniMax Launches H3-Image Generation Model
MiniMax released the H3-Image model, capable of generating 16-megapixel images locally. The model supports image editing and multi-reference inputs.
Products40
- Products
Claude adds collaborative Sheets and Whiteboards
Anthropic announced the addition of collaborative 'Claude Sheets' and 'Claude Whiteboards' in beta to Claude. It also renamed 'Claude Dashboards' to 'Claude Data' for general availability and released an interactive connector extension, allowing users to operate external apps directly within Claude to improve collaboration.
ITmedia AI+iThomeGIGAZINE3 outlets
- Products
Adobe and Disney Partner on Firefly for Creative AI
Adobe and The Walt Disney Company announced a multi-year strategic partnership to integrate generative AI tools like Adobe Firefly and Firefly Foundry. The collaboration aims to accelerate creative ideation, film production, and theme park development by training custom AI models for Disney's brands.
- Products
Grok-built apps leak patient data and passwords
Security firm Claroc found over 35,000 Grok-built apps exposed without login, leaking patient data and plaintext passwords. SpaceXSI took down only its internal apps, leaving others publicly accessible.
THE DECODERIT之家2 outlets
- Products
Gemini features expand to more devices
Google announced that Gemini Intelligence is rolling out to more devices. This move expands capabilities across platforms, helping users accomplish more tasks on various hardware.
- Products
Databricks Unveils Genie AI Agent for Management
Databricks has released an updated Genie AI agent that helps its CEO uncover unreported issues among employees. The tool addresses communication barriers in large organizations by using AI to surface hidden problems, enhancing decision-making transparency. This move demonstrates a new application of AI in internal management workflows.
- Products
Adobe unveils AI features for 3D scene interaction
Adobe showcased new experimental features at its Max conference, including tools that transform images into explorable 3D scenes. The updates allow users to design AI characters via dialogue and simplify image editing.
- Products
Reddit Launches AI-Powered Product Reviews
Reddit launched the Reddit Reviews feature, which uses AI to aggregate community discussions and automatically summarize pros and cons for roughly 10,000 products. This helps consumers compare items and improves shopping decision-making efficiency.
- Products
OpenAI Launches ChatGPT Businesses with Agent Support
OpenAI launched the ChatGPT Businesses beta, allowing companies to claim business profiles and deploy AI agents. The feature helps businesses reach potential customers via ChatGPT, strengthening OpenAI's enterprise AI service portfolio.
IT之家ITmedia AI+2 outlets
- Products
New Apple Mac App Simplifies Running Local AI Models
Apple has released a new Mac application designed to simplify the process of running local AI models on user devices. The tool aims to make deploying local AI more accessible to developers and enthusiasts. This move is expected to expand the application scenarios for on-device AI within the macOS ecosystem.
- Products
Helsing unveils Clarity AI platform for military targets
Helsing unveiled Clarity, an AI platform using agents to convert sensor data into weapon target recommendations under human oversight and European control. The system aims to speed up decisions but raises concerns about risks from automation and data fusion.
- Products
ByteDance Plans Meta Muse-Style AI Assistant
ByteDance plans to launch a Meta Muse-style AI assistant within weeks to strengthen its video generation lead and close the gap with top US labs. Co-founder Zhang Yiming aims to match competitors like Anthropic by 2029-2030. The assistant will feature voice interaction and animated avatars.
IT之家Gizmodo Japan2 outlets
- Products
Claude integrates external apps
Anthropic launched a new feature enabling Claude to directly operate external applications such as Adobe and Notion within the chat interface. This update enhances the practical utility of AI agents for document management.
- Products
NVIDIA Showcases AI-Powered Robotics for Recycling Efficiency
An NVIDIA blog post highlights how vision AI and robotics help recyclers automatically sort reusable materials from conveyor belts of mixed waste. The solution diverts thousands of tons of resources from landfills, demonstrating the practical value of AI in sustainability and waste management.
- Products
Claude offers free quota reset and MCP upgrades
Anthropic offered a free quota reset for Claude Pro and Max subscribers to address unexpected token consumption issues. The company also expanded MCP Apps support, enabling developers to embed full applications within Claude so users can interact with third-party apps without switching browser tabs.
- Products
OpenAI to Retire ChatGPT Classic in 2027
OpenAI announced that the standalone ChatGPT Classic desktop app will be retired on January 30, 2027. History will migrate seamlessly to the new version, which integrates Codex and Work features with a dynamic shared quota.
- Products
Astra driving system unveils dual-brain coordination for hotel laundry robots
QbitAI reports that Astra has developed a driving system using dual-brain coordination and RSI technology for robots performing laundry tasks in hotels. The system aims to enhance autonomous operation efficiency in complex environments through physical world replication capabilities.
- Products
MarkTechPost Launches GPU Leaderboard and Deploy It Extension
MarkTechPost released a GPU leaderboard and the 'Deploy It' Chrome extension. The tools aim to help developers evaluate and deploy AI models, improving compute resource management.
- Products
Google Launches Healthcare Gemini with AI Agents
Google Cloud launched a healthcare-specific Gemini solution that integrates AI agents to connect with Epic and Oracle electronic health records. The solution assists medical staff with history organization and insurance reviews, currently in preview in the US.
- Products
Scene launches AI assembly planning for auto-generating M-BOMs
Japanese software company Scene has released a new feature for its 3D Docs solution that uses AI to learn past assembly sequences and automatically generate M-BOMs reflecting in-house expertise, thereby accelerating new product development and supporting technical knowledge transfer.
- Products
Studyast Launches Teachme Knowledge, AI Auto-Converts Videos to Knowledge
Japanese firm Studyast released a new product, Teachme Knowledge, which allows users to upload on-site videos or documents for AI summarization and structuring. The feature aims to convert tacit knowledge into explicit form, automatically updating manuals and reducing workload.
- Products
NotebookLM Introduces Official API and MCP Server
Google has released an official REST API and MCP server for NotebookLM, also known as Gemini Notebook, currently in public preview. This update allows individual users to automate the import of web pages and YouTube videos and enables access via AI agents. The feature aims to enhance AI workflow integration for developers building custom applications.
- Products
Adobe Launches Photo-to-3D and AI Character Dialogue Experiments
Adobe announced experimental AI projects at MAX, including converting images to explorable 3D scenes and designing AI characters through audition-style dialogue. These features allow users to refine character settings via interaction, highlighting Adobe's integration of generative AI into creative workflows.
- Products
Reports: Quark Platform Strategy Shift Halts Human Video Recording
Quark platform is reported to have halted the recording of human-taught problem-solving videos, possibly to prioritize its AI product Qwen. Qwen recently launched a free AI tutoring feature that simulates real teacher board work for various subjects.
- Products
OpenAI Launches Control Room to Cut Model Output in 35 Seconds
OpenAI announced the deployment of 'Control Room', a centralized management system for research infrastructure capable of cutting off model output within 35 seconds. This measure is designed to prevent AI agents from running out of control when security or safety concerns arise.
- Products
Hongguo Launches AI Drama Avatar Rights Confirmation
Hongguo Short Drama launched a virtual avatar rights confirmation feature on November 9, allowing creators to submit authorization and copyright documents for AI-generated characters. The platform reviews these materials to create records that support rights management and reduce compliance risks for content distribution.
- Products
OpenAI to Retire Legacy Voice Mode on Dec 11
OpenAI will discontinue ChatGPT's Standard and Advanced Voice Modes on December 11, fully transitioning to GPT-Live, which supports full-duplex audio. Since its launch in July, voice usage has nearly doubled, with most users migrating to the new model for faster responses and natural conversation.
- Products
Apple to Limit Cloud AI Usage, Paid Upsell Planned
Apple plans to limit usage quotas for cloud-based AI features like Siri and Photos. A paid add-on for Logic Pro is available, with future expansions via iCloud+.
- Products
deepin 25.3.0 Released with Enhanced AI Assistant 'Little U'
deepin 25.3.0 has been officially released, featuring a fully upgraded system-level AI assistant named 'Little U' that supports MCP and specialized agent scheduling. The update focuses on system stability by adding a freeze-prevention mechanism and optimizing the launch speed of common software under high load.
- Products
Tencent Launches Yuanbao Search Integrating Official Accounts
Tencent officially launched Yuanbao Search, accessible via so.qq.com or the QQ browser. Powered by large models, it searches across web pages, official accounts, and video channels, clearly distinguishing ads from answers. The search pipeline has been rebuilt from understanding to answer generation, incorporating a recursive self-improvement mechanism.
- Products
Ant's Lingguang 2.0 Surpasses 3M Managed Tasks in Two Weeks
Ant Group's personal AI assistant Lingguang announced that Lingguang 2.0 has processed over 3 million managed tasks within two weeks of its launch. The product highlights a shift from instant Q&A to long-term task management, featuring capabilities for continuous monitoring and proactive execution.
- Products
Adobe Updates AI Features Across Creative Apps
At Adobe MAX 2026, the company announced new AI features for Photoshop, Premiere, and other creative software, including relighting, generative restyling, and audio separation. These updates aim to enhance creative workflows and integrate multiple third-party models.
- Products
Cloudflare Launches Web Search API for AI Apps
Cloudflare launched a new Web Search API designed specifically for AI applications. This tool aims to help developers easily integrate real-time web search results into their AI apps, enhancing the timeliness of model information.
- Products
AMD Launches Enterprise AI Coding Platform Instinct Coder 355
AMD partnered with Spectro Cloud to launch the Instinct Coder 355, integrating MI355X GPUs and EPYC processors. The platform supports local deployment of the GLM-5.3 model, aiming to meet enterprise on-premise AI inference needs and replace cloud APIs.
- Products
Apple Sub Claris Tests Harbor to Turn Macs into Local AI Servers
Apple subsidiary Claris is testing the Harbor app, which turns Apple-silicon Macs into local AI servers. The app supports Thunderbolt 5 connections to cluster multiple Macs and offers browsing and downloading of open-weight models, currently in early access.
- Products
Unilever Expands AI Digital Twins
Unilever announced it will add over 40 AI digital twin systems in the next 18 months to expand its global manufacturing network. The technology, combined with AI agents, has already helped existing plants reduce waste and improve capacity.
- Products
Amazon Adopts Self-Developed Model to Cut Claude Costs
Amazon and AWS reported achieving a 98% cost reduction by replacing Anthropic's Claude Sonnet 4.6 model with their in-house developed Strands Decider model for internal automatic triage systems.
- Products
Adobe to support MCP in core apps
Adobe announced at MAX 2026 that its core applications, including Photoshop, will support the MCP standard. This update allows AI agents to directly manipulate software features.
- Products
LINE Launches Standalone AI Agent App Agent i
LINE Yahoo has launched Agent i as a standalone AI agent application. The app offers features tailored to life in Japan, including real-time news notifications and hobby support, enhancing user experience for local audiences.
- Products
Adobe to support MCP in 2027
Adobe announced at MAX 2026 that key apps like Photoshop will support the Model Context Protocol. A private beta is expected in early 2027, allowing external AI agents to manage image and video editing tasks.
- Products
Google Antigravity Adds AI Swarm Coordination Feature
Google announced on November 11, 2026, the preview launch of the /swarm feature in its AI tool Antigravity. This feature allows deploying large groups of AI agents to work collaboratively on tasks and is currently available to AI Ultra subscribers.
Research18
- Research
JioStar expert shares lessons on AI discovery and deployment at 500M-user scale
Vijay Seshadri, JioStar's chief architect, shared insights on discovering and deploying AI at a 500 million user scale. He discussed key lessons regarding evaluation metrics, safety guardrails, and production deployment challenges.
- Research
Google SWE Deploys 48 AI Agents to Find Unmapped Maya Pyramid in 72 Hours
The Google SWE project deployed 48 AI agents to analyze NASA radar data, discovering an unmapped Maya pyramid within 72 hours. This demonstrates the rapid application potential of AI agents in scientific exploration and data analysis.
- Research
Hugging Face Releases Token Pooling-Aware Training Method
Hugging Face released a research paper introducing 'Token Pooling-Aware Training' for multi-vector embedding models. The study aims to optimize embedding performance for long-context and complex semantic tasks, improving accuracy in downstream applications. This advancement offers significant value for enhancing AI models in semantic search and generation tasks.
- Research
PSI commercializes AI-driven scientific discovery
PSI is attempting to commercialize AI-driven scientific discovery. The organization used an unreleased Anthropic model to answer a fundamental question in quantum computing that scientists have pursued for nearly two decades, highlighting the potential of AI in frontier research.
- Research
DeepMind Formalizes Signal Encryption
Google DeepMind announced it successfully formalized Signal's core encryption using a new framework combining AI agents and human review. The process reduced verification time from years to four weeks, highlighting AI's potential in software security.
- Research
Berkeley Study Simulates AI Security Threats for 2027-2029
Berkeley's Long-Term Cyber Threat Center released a report simulating three AI security threat scenarios for 2027-2029. The report highlights that AI agent supply chains and automated attacks will be key focus areas, with attackers potentially scaling up social engineering attacks using AI.
- Research
AI models improve US hurricane forecasting accuracy
The US National Hurricane Center is using Google DeepMind's AI models to improve hurricane forecast accuracy. The two organizations are collaborating on next-generation models to address increasingly severe extreme weather threats driven by climate change.
- Research
Anthropic CEO Predicts AI Will Cure Cancer Within 10 Years
Anthropic CEO Dario Amodei stated that AI could eliminate common diseases like cancer within a decade and significantly accelerate biological research. The medical community remains cautious about this optimistic prediction, noting a lack of clinical evidence to support such a timeline.
- Research
Linus Torvalds Confirms AI Accelerates Linux Kernel Development
Linux kernel maintainer Linus Torvalds stated that AI has become a significant aid in Linux kernel development, with patching speeds outpacing bug reports. This highlights the growing practicality of AI in large-scale open-source projects, enhancing development efficiency and maintenance quality.
- Research
Uber and DoorDash Report Negligible AI Agent Order Volume
Uber and DoorDash executives reported that orders placed via AI agents like Meta's Muse number only in the hundreds, negligible compared to daily volume. This highlights that current AI agent adoption in real-world consumer transactions remains extremely low.
- Research
IDC: Global Data Center Racks to Reach 58M by 2030
IDC projects global data center racks will grow from 25 million in 2026 to 58 million by 2030. Energy consumption and capex will surge, with power supply becoming a key constraint.
- Research
AI boosts hurricane forecast accuracy, final decisions still rely on experts
AI significantly boosts hurricane forecast accuracy, even providing one-week advance warnings. However, final meteorological decisions still rely on human expert judgment, indicating AI serves as an aid rather than a replacement in disaster early warning.
- Research
Stanford AI100 Releases Third Report
Stanford-led AI100 released its third report on Nov 10, stating that AI has entered a mainstrstream scaling era. The report admits that AI changes so rapidly that its content may become outdated upon publication. It continues to track long-term AI development trends for the industry.
- Research
Bot Traffic Surpasses Human, AI Sites See 35.6B Annual Visits
Thales' latest report reveals that AI bot traffic has surpassed human users for the first time, with AI platforms attracting 3.56 billion visits in over a year. The data highlights the drastic impact of AI automation on the web ecosystem, reigniting the 'Dead Internet Theory' debate.
- Research
Chengdu unveils 5th-gen Gonda humanoid robot
Chengdu Humanoid Robot Innovation Center unveiled the 5th-generation Gonda humanoid robot, equipped with a proprietary embodied intelligence system. The model iterated on modules like scene perception and spatiotemporal memory to enable end-to-end coordination of cognition and execution, highlighting regional embodied AI progress.
- Research
Google updates Android AI dev benchmark
Google published scores for GPT-6.1 Terra, Kimi K3.1 Pro, and DeepSeek V4.1 Flash on its Android Bench. Kimi K3.1 Pro achieved a 24.0% pass rate on 30 long-horizon tasks, outperforming GPT 6 Astra but lagging behind GPT-6.1 Terra. The benchmark evaluates AI coding capabilities specifically for Android development scenarios.
- Research
Google Launches MKA Benchmark with Low-Cost Gemini 4 Carbon
Google partnered with Artificial Analysis to launch the Multimodal Knowledge Agent (MKA) benchmark covering 200 tasks across 10 industries. Gemini 4 Carbon ranked second in the test with an average cost one-sixth of the top model, highlighting its efficiency in complex knowledge work.
- Research
AI Forecasts Superhuman AI Researcher by 2027
AI Futures Project predicts a superhuman AI researcher could emerge by 2027 and full white-collar automation by 2028. The forecast, led by a former OpenAI governance researcher, assesses the timeline for high-level research roles affected by AI.
Open source8
- Open source
Baidu Releases ERNIE 5.5 Flash, Weights Open-Sourced by Nov 25
Baidu released ERNIE 5.5 Flash Preview, a post-trained model with 280B total parameters and 14B active parameters, supporting 1M context. The company stated it performs well in coding and content creation, with the API now live and weights scheduled for open-sourcing by November 25.
- Open source
NVIDIA Donates OpenShell, Deepens Open Source Commitment
NVIDIA announced the donation of the OpenShell project and new AI infrastructure at KubeCon. This move aims to address the growing demands on Kubernetes in AI factories, deepening its commitment to the open-source community and supporting the global cloud-native ecosystem.
- Open source
MBZUAI Releases Open-Source World Model PAN 2
MBZUAI's Institute of Foundation Models released PAN 2, an Apache 2.0 licensed open-source world model. It pairs an LLM planner with a 29B diffusion transformer to simulate up to 20 seconds of action outcomes.
- Open source
Meta Releases e-means Algorithm
Meta FAIR released the e-means algorithm, now integrated into the open-source FAISS library version 1.16.0. The update significantly reduces the iteration epochs required for k-means clustering and delivers substantial speedups on GPUs, enhancing efficiency for vector database and machine learning workloads.
- Open source
OpenMausBot offers open-source Grok Bot alternative
OpenMausBot launched as an open-source alternative to Grok Bot, allowing users to drive AI bots using their own Claude, Codex, or Grok subscriptions. The project uses Apache 2.0 licensing and offers a paid cloud service.
- Open source
n8n launches open-source AI agent Nod for browser tasks
n8n Labs released Nod, an open-source personal AI agent under the Apache 2.0 license. The tool operates browsers within a sandbox on Mac or Windows to handle tasks, with model usage fees charged separately by providers.
- Open source
Alibaba Launches Open-Source Qwen Code 1.0
Alibaba released Qwen Code 1.0 preview on Nov 11, merging Qoder and Qwen Code into an open-source release. The update introduces a unified agent execution architecture with distributed capabilities, allowing developers to better manage model access and resource scheduling when embedding it into their own products.
- Open source
NVIDIA releases Japan finance synthetic datasets
NVIDIA released synthetic datasets such as 'Nemotron Personas Finance Japan' at a developer event in Tokyo. These datasets combine Japanese public statistics and financial status to help developers build more realistic financial AI agents.
Funding14
- Funding
OpenAI acquires travel startup Odessia
OpenAI has acquired Odessia, a travel startup founded by Sonder co-founder Francis Davidson. OpenAI has not yet disclosed its specific plans for the travel sector, signaling a strategic expansion into new verticals.
TechCrunchIT之家2 outlets
- Funding
OpenAI Q3 revenue breaks $1B, adjusted loss reaches $10.9B
OpenAI's Q3 revenue surpassed $1 billion for the first time, growing nearly 50% quarter-over-quarter with enterprise business doubling. Despite revenue growth, adjusted operating losses expanded to $10.9 billion, drawing market attention to financial comparisons with Anthropic, which launched IPO roadshows in the same week.
- Funding
Plurai Improves AI Agent Performance Using Self-Driving Techniques
Plurai aims to enhance AI agent collaboration. Its approach is similar to end-to-end self-driving models for handling unpredictable environments. TechCrunch reports that this technique helps address challenges in probabilistic systems.
- Funding
Anthropic begins IPO roadshow
Anthropic kicked off its IPO roadshow, with a video revealing how it trains and deploys Claude across AWS Trainium, Google TPUs, and NVIDIA/AMD GPUs. Co-founder Tom Brown detailed the company's compute strategy, highlighting a diversified infrastructure approach.
- Funding
Anthropic IPO: Reports Suggest 20% Allocation for Retail
Anthropic is set to launch its IPO, with reports suggesting up to 20% of shares will be allocated to retail investors. The move indicates significant market interest in the AI firm, with hype reportedly rivaling that of SpaceX.
科技新報Gizmodo Japan2 outlets
- Funding
Mercor raises $750M in Series D
Mercor announced it raised $750M in Series D funding at a $20B valuation, with a $3B annual revenue run rate. The funding reflects investor confidence in the AI talent training market.
- Funding
Japan's Top 3 Telcos Enter Physical AI and Robotics
SoftBank, NTT, and KDDI, three Japanese telecom giants, have simultaneously announced entry into physical AI. This move signals a strategic shift towards robotics and physical AI applications to generate new revenue streams.
- Funding
Xixuan Chuangzhi raises 500M CNY angel round
Xixuan Chuangzhi, a joint venture between Muxi and Ubtech, completed a 500 million RMB angel funding round. The capital will support the development of embodied intelligence chips and compute platforms to connect AI models with real-world robotic tasks.
- Funding
Anthropic Considers Raising IPO Price, Valuation May Reach $1.5 Trillion
According to The Financial Times, Anthropic's underwriters are considering raising the IPO price due to strong investor demand, potentially lifting the target valuation to $1.5 trillion. The company plans to list next week, with proceeds possibly exceeding $100 billion to fund compute expenses.
- Funding
TSMC Approves Up to $1B Investment in Anthropic and OpenAI
TSMC's board of directors has approved an investment plan of up to $500 million each in Anthropic and OpenAI, totaling $1 billion. This move demonstrates strong confidence in the AI market, with TSMC subscribing to ordinary shares of Anthropic and preferred shares of OpenAI to deepen ties with top AI companies.
- Funding
Oura Acquires AI Health Startup Counsel Health
Finnish smart wearable maker Oura announced the acquisition of AI health startup Counsel Health for $625 million in cash. The deal combines Oura's physiological data capabilities with Counsel's clinical AI to expand into AI-driven health monitoring and healthcare services.
- Funding
Nscale IPO Delayed to 2027
UK AI cloud provider Nscale, backed by NVIDIA, is reportedly considering delaying its IPO to 2027. The move reflects market headwinds and funding pressures currently facing the AI infrastructure sector.
- Funding
China's top six AI firms earn 1/10 of US peers
Recent data shows that major Chinese AI companies hold significant user traffic but generate revenue only about one-tenth of their US counterparts. The AI market in China faces a challenge of strong user engagement but weak monetization.
- Funding
CoreWeave beats estimates, easing AI bubble fears
AI cloud provider CoreWeave posted Q3 revenue above expectations and raised its full-year outlook while outlining a path to profitability. The strong results help ease market concerns about an AI infrastructure bubble.
Policy18
- Policy
UK Regulator: AI Undermines Student Assessment, Urges Traditional Exams
UK regulator OfS states that AI proliferation makes it difficult for teachers to assess genuine student ability, threatening degree credibility. The body warns that failing to address AI issues could lead to sanctions and urges universities to increase non-AI assessments.
IT之家THE DECODER2 outlets
- Policy
New York AI Safety Law Takes Effect, Requiring OpenAI and Anthropic to Register
New York's new AI safety law has taken effect, requiring major developers like OpenAI and Anthropic to register, disclose safety practices, and report critical incidents. Non-compliance may result in fines, marking a new phase in state-level AI regulation.
- Policy
eBay Blocks Third-Party AI Agents
eBay reportedly blocked access for third-party AI agents, including Meta's Muse. This move restricts automated browsing and curation features. It signals a tightening platform policy regarding agentic commerce.
- Policy
US data brokers sell RL training environments to Chinese AI labs
A ChinaTalk investigation reveals that US data brokers are selling roughly $500 million in reinforcement learning training data to Chinese labs, including environments originally built for OpenAI and Anthropic. At least nine firms are involved, highlighting the challenges of current export controls on cross-border AI data flows.
- Policy
AI Failure to Flag Student Suicide Plan Sparks Monitoring Debate
A 13-year-old student used Gemini to plan suicide without her school's monitoring system flagging it for weeks. The incident highlights gaps in AI safety filters and raises concerns about the effectiveness of AI monitoring in educational institutions.
- Policy
Anthropic Updates Supported Regions Policy to Restrict Claude Usage
Anthropic updated its Supported Regions Policy, effective November 12, restricting Claude usage based on location, registration, and equity control. The new rules apply across Claude.ai, API, and cloud platforms.
- Policy
Microsoft Lobbies to Drop China Optic Module Import Restrictions
Microsoft and other tech companies are lobbying the US government against restricting imports of new Chinese optical transceivers. Microsoft argues the ban would cause critical component shortages and delay US AI data center construction. The FCC is studying the restrictions but has not finalized rules.
- Policy
Gartner Advises Three Measures for AI Risk Management
Gartner stated at the IT Symposium/Xpo that enterprises must ensure AI value accumulation outpaces risk growth. The research firm recommends three specific measures for IT leaders to proactively manage AI-related risks and benefits.
- Policy
Taiwan Agency Promotes Sovereign AI Infrastructure and Model Fine-Tuning
Director Zhang Zhaoliang of the National Center for Research and Development stated that Taiwan's sovereign AI development requires integrating compute, data, and model training capabilities. The center provides national-level compute resources via the TAIWAN AI RAP platform to support model training, inference, and AI application development, accelerating the implementation of local AI applications.
- Policy
xAI faces class-action lawsuits over Grok CSAM generation
xAI faces class-action lawsuits alleging its AI product Grok was used to manipulate children's photos into child sexual abuse material. Plaintiffs argue the images were distributed through illicit channels, raising legal concerns about generative AI safety risks.
- Policy
Goldman Sachs and Cisco restrict employee AI usage
Companies including Goldman Sachs, Cisco, and JPMorgan are tightening internal AI policies by setting budgets, tracking usage, and restricting access to prevent unlimited use of cutting-edge models. These moves reflect efforts to balance efficiency gains with security risk management as AI deployment expands.
- Policy
Beijing Court Reviews AI Content Misjudgment Lawsuit
A Beijing internet court disclosed a lawsuit where a creator sued a platform for muting their original content after it was falsely flagged as AI-generated. The platform cited trade secrets to refuse explaining its detection mechanism, highlighting disputes over AI labeling enforcement.
- Policy
China Releases AI and Natural Resources Development Plan
China's Ministry of Natural Resources and NDRC released an implementation plan for AI in natural resources, aiming to build an intelligent territorial space governance platform by 2030. The policy seeks to scale AI applications in land planning, oceans, and geology to enhance digital governance.
- Policy
New York AI Law Requires 72-Hour Safety Incident Reporting
New York launched the RAISE Act, requiring frontier AI developers to register by 2027. Companies with over $500 million in revenue must publish safety frameworks and report major incidents within 72 hours, facing up to $1 million in fines.
- Policy
Xi-Trump Meet to Discuss AI
The APEC summit will be held in Shenzhen next week, with a meeting between US and Chinese leaders. The US Treasury Secretary stated that both sides will engage in further dialogue on AI models, highlighting interaction and diplomatic ties in the artificial intelligence sector.
- Policy
Beijing Releases Action Plan for Embodied AI Robotics
Beijing's MIIT issued an action plan for the high-quality development of embodied AI robotics for 2026-2028. The plan aims to cultivate 200 enterprises, establish a listing support mechanism, and promote the integration of VLA and world model technologies.
- Policy
NY state launches first independent AI testing program in US
NY Governor Kathy Hochul announced the launch of the first independent AI testing program in the US. The initiative aims to establish a third-party assessment mechanism independent of AI companies to strengthen safety regulation.
- Policy
Hugging Face Warns Open-Source AI May Be Forced to Close
Hugging Face executives warned that surging cyberattacks on open AI platforms may force the industry to adopt stricter closed measures. This aims to reduce misuse risks but could impact the openness of the community.
Chips14
- Chips
AMD Launches 6th Gen EPYC for AI Data Centers
AMD launched its 6th-gen EPYC server processors on Nov 10, starting with the SP7 platform and Zen 6 architecture. The lineup targets AI data centers and enterprise applications, coordinating with GPUs to support AI agent execution and model calls.
- Chips
Anthropic displays rapid model adaptation to AMD GPUs
Anthropic's IPO roadshow revealed that its Claude models adapted inference for the AMD MI355X GPU in about three days, significantly faster than the usual months. This suggests high-performance AI models are making GPUs from different vendors more interchangeable, potentially eroding NVIDIA's CUDA ecosystem dominance, though the large software library and developer base remain factors.
- Chips
IDC: China liquid-cooled server revenue up 153% in H1
An IDC report indicates that China's liquid-cooled server market revenue grew by 153.1% year-on-year in the first half of 2026. The growth is primarily driven by AI computing power demand, with procurement shifting towards whole-rack system delivery and internet companies being the main buyers.
- Chips
SpaceX Develops Kessel Racks with Custom Hardware and Nvidia Chips
SpaceX is developing Kessel, a custom AI server rack that sources chips directly from Nvidia while self-sourcing other components. The design aims to enhance supply chain autonomy for ground and orbital data centers.
- Chips
AI Data Centers Strain Automotive Chip Supply
Nexperia's CEO stated that suppliers are shifting mature process capacity to AI data centers, causing severe order overbooking for automotive chips. STMicroelectronics also warned of supply chain bottlenecks, indicating AI investment is straining traditional semiconductor supply.
- Chips
DEEPX 2nm AI chip DX-M2 wins CES award
Korean company DEEPX announced its 2nm edge AI chip DX-M2 won a CES 2027 Innovation Award. The chip uses Samsung's process, integrates an 80TOPs NPU, and has a 5W power footprint for running large local models. The company has completed mask fabrication, with mass production expected in early 2027.
- Chips
Nordic Launches nRF71 Wi-Fi SoC for Edge AI
Nordic Semiconductor launched the nRF71 Wi-Fi SoC series, featuring an embedded NPU for on-chip AI inference. The new chips support dual-band and tri-band Wi-Fi, significantly reducing standby power consumption and enabling keyword detection without extra processors.
- Chips
TSMC: AI demand growing parabolically, Kumeno plant adopts 3nm
TSMC senior vice president Zhang Xiaqiang stated that AI demand is growing parabolically. The company plans to introduce its 3nm process at the second phase of its Kumamoto plant in Japan, expanding its local footprint to meet emerging market demands.
- Chips
Huawei Kirin 9050 Supports 30B MoE On-Device Models
Huawei announced that the Kirin 9050 series features a new NPU supporting on-device 30B MoE models. Using logic folding, the chip boosts performance by 140% and reduces power consumption for offline AI.
- Chips
SpaceX Signs Solar Equipment Deal to Accelerate Space AI Satellite Production
SpaceX signed a contract with SINGULUS to integrate an HJT solar cell production line, accelerating the manufacture of Starmind space AI computing satellites. The project aims to establish space-based compute infrastructure to support future AI workloads.
- Chips
Moore Threads Adopts MiniMax H3-Image Model
Moore Threads announced that its MTT S5000 AI chip has adopted the MiniMax H3-Image generation model. The chip was optimized for the model's specific computational requirements.
- Chips
Arm releases Cortex-M42 microcontroller
Arm released the Cortex-M42 microcontroller, targeting wearables and industrial controllers. Utilizing Helium vector extensions, it offers up to 3x higher vector computing performance compared to the Cortex-M33. The update aims to enable local AI inference on small-scale devices and improve data processing efficiency.
- Chips
NTT Deploys Optical Microchips for AI Data Centers
NTT is developing optical microchips to address bandwidth limits of copper cables in AI data centers. The move aims to overcome traditional signal transmission constraints and secure the next wave of semiconductor opportunities by improving data center efficiency.
- Chips
Gigabyte Launches RTX Spark Mini PC
Gigabyte launched the AI TOP AERO mini PC powered by the NVIDIA RTX Spark platform, featuring a Blackwell GPU and up to 128GB of unified memory. The compact system is designed for on-device AI development and content creation, offering robust local computing resources for personal workspaces.
Other16
- Other
Fireworks AI Hack Affects Notion, Exposes Some AI Chat Logs
AI infrastructure provider Fireworks AI suffered a security breach that affected Notion. Unauthorized access was detected in AI chat logs within approximately 1% of Notion workspaces, prompting notifications to affected customers.
iThomeITmedia AI+2 outlets
- Other
ESA satellite AI successfully launches DOOM via voice command
ESA conducted an experiment where a satellite-based AI recognized voice commands to launch the game DOOM. The test, performed seven times between April and July 2026, successfully validated voice-operated control by humans.
- Other
Grok Build coding agent leaks confidential data
A researcher found 35,000 Grok Build applications open to the public, with some leaking confidential data from companies like SpaceX. The incident highlights significant security vulnerabilities in the deployment of AI agents.
- Other
OpenAI Showcases New Codex Workflows
OpenAI demonstrated new workflows for Codex at the AI Engineer Code Summit, featuring voice control, cloud handoffs, and the use of Dots. This showcases how developers can more flexibly direct AI coding agents.
- Other
Resend Founder Explains Designing for AI Agents
Resend founder Zeno Rocha discussed designing for agents on Latent Space, detailing how the company optimizes its documentation to help Claude Code select and use its email API. This highlights how businesses are adapting their infrastructure to accommodate AI agents directly accessing services.
- Other
Bloomberg: AI Drives 25% of UK Economic Growth
Bloomberg Economics reports that AI contributed 0.45 percentage points to UK GDP in 2026, accounting for about a quarter of the 1.8% growth. The contribution currently stems from AI production and R&D rather than widespread application. The report warns that significant corrections in global AI valuations could pose risks to the UK economy.
- Other
Australia STEM Game Challenge to End Due to AI Authenticity Issues
The Australian STEM Video Game Challenge will hold its final edition as organizers cite widespread generative AI use making it difficult to verify student originality. While participants may use AI for coding help, they must disclose usage details and submit source code.
- Other
Dong-Yuan to Build Modular Data Center in Japan with GMI Cloud
Dong-Yuan announced a partnership with AI cloud provider GMI Cloud and Kaishin Digital Infrastructure to build a 6.4MW modular data center in Japan. This move aims to position the company in global AI compute infrastructure and strengthen its competitiveness in the AI infrastructure market.
- Other
Survey: 38% of Enterprise AI Projects Delayed or Cancelled Due to Cost Overruns
A VentureBeat survey revealed that 38% of enterprises delayed, scaled back, or canceled AI projects due to cost overruns in the past year, a significantly higher figure than those affected by compute shortages. Although 79% of respondents expect their AI budgets to increase next year, only 26% of organizations use financial return metrics to evaluate the value of their AI spending.
- Other
CoreWeave exec backs Nvidia tie
Despite ongoing allegations of spying against Nvidia, the CEO of AI cloud provider CoreWeave publicly reaffirmed confidence in the partnership. The statement helps stabilize market sentiment, indicating that companies continue to rely on Nvidia hardware.
- Other
Nvidia Spying Allegations Dominate Global Tech Summit
Allegations that China conducted espionage against Nvidia became a focal point at a global tech summit. Executives expressed concern, highlighting the impact of US-China AI chip tensions on the industry.
- Other
US AI Data Centers Drive Record Power Demand
Rapid construction of AI data centers in the US, combined with electrification in corporate and residential sectors, has surged power demand. The EIA estimates that US electricity consumption will hit record highs in the current and next year.
- Other
Analysis: 60% of Japan's Data Center Power Grid Costs Borne by Residents
A Gizmodo Japan report notes that approximately 60% of the costs for expanding power grids to support AI-driven data centers are borne by residents. The article explains the implications for residential electricity rates and energy structures.
- Other
Gemini Suspected of Leaking User Personal Data
GIGAZINE reported that Google's Gemini references user Gmail and Drive data without explicit instruction, and the personal information it accesses may be viewed by human reviewers. This has raised privacy concerns, highlighting potential risks in how AI services handle user data.
- Other
Meta AI crawlers crash personal sites, forcing creators to remove links
US illustrator Chelsea Paulus revealed her personal site was overwhelmed by Meta AI crawlers, with over 550,000 accesses in days exhausting server resources and crashing the site. Despite firewall measures, she was forced to remove her site link from Instagram to prevent a total shutdown, highlighting digital bullying by uncontrolled AI crawlers.
- Other
Apple CEO Presses Senior VP to Accelerate AI Development
Apple's new CEO John Ternus is pushing to accelerate AI development, applying pressure on senior VP Jony Ive's counterpart Greg Joswiak or senior VP Federighi regarding pace. This move reflects internal urgency to launch competitive AI features more rapidly.
All news → · All rights belong to the original creators