AI news, Nov 13, 2026: Gemini 4 Hydrogen, Huang Departure, ChatGPT Wallet
Google DeepMind released Gemini 4 Hydrogen, which offers a 2.5x speed improvement over previous versions and is expected to be integrated into Google Search. NVIDIA CEO Jensen Huang announced his departure, with Colette Kress serving as interim CEO, raising market concerns about the AI supply chain. Concurrently, OpenAI launched the ChatGPT Wallet feature, enabling approved purchases by AI agents, which enhances commercial transaction capabilities.
Lead written by AI from the day’s stories
Top stories
- Models
Google DeepMind Introduces Gemini 4 Hydrogen
Google DeepMind has introduced the new Gemini 4 Hydrogen model. The model aims to enhance visual and intuitive responses and is slated for global integration into Google Search to provide more helpful assistance.
Google AIGoogle DeepMindHacker News11 outlets
- Other
Jensen Huang Resigns from NVIDIA, Colette Kress Takes Over
Jensen Huang, who led NVIDIA for 33 years, has resigned. Colette Kress takes over as interim CEO. The move sparks concerns about American AI leadership and the open-source ecosystem.
量子位Latent Space科技新報7 outlets
- Products
OpenAI Launches ChatGPT Wallet
OpenAI introduced ChatGPT Wallet, enabling agents in ChatGPT Work and Codex to complete purchases with user approval. The feature integrates existing card protections, enhancing the commercial transaction capabilities of AI agents.
OpenAIIT之家TechCrunch5 outlets
- Models
Google Announces Gemini 4 Hydrogen, Free and 2.5x Faster
Google announced the free AI model Gemini 4 Hydrogen on November 12. The model runs 2.5 times faster than its predecessor and outperforms competitors, integrating into the Gemini app and search.
- Chips
Chinese AI Firms Suspected of Using NVIDIA Blackwell Chips for Model Development
AI research group Unemployment Labs analyzed multiple Chinese AI models and detected traces of training on restricted NVIDIA Blackwell GPUs. The findings suggest Chinese AI companies continue to use high-end chips for model development despite export controls, highlighting challenges in enforcing chip regulations.
GIGAZINETHE DECODERTechCrunch3 outlets
Models12
- Models
LLMON: New LLM-Native Markup Language for Granite 4.1 Models
Hugging Face has introduced LLMON, a markup language designed specifically for large language models. It is now supported in Granite 4.1 3B and 8B models, aiming to improve how LLMs handle structured data.
- Models
Edge0-145B Model Enables Local Execution
A Hacker News discussion highlights that the Edge0-145B model runs frontier AI capabilities on personal computers. This indicates large models are advancing toward edge computing, enabling users to access advanced AI locally without the cloud.
- Models
Sakana AI Begins In-House Model Development
Sakana AI has announced it is starting to develop its own AI models from scratch, a shift from its previous focus on utilizing existing models. This move marks a new strategic direction toward fundamental model research and development.
- Models
Liquid AI releases LiquidCodec streaming audio codec
Liquid AI released LiquidCodec, a 74M-parameter streaming audio codec powering LFM3-Audio-2B. It runs at 12.5 Hz and encodes 361x faster than real-time on a single CPU thread, enhancing multimodal model performance.
- Models
NVIDIA releases synthetic datasets for Japanese AI
At the 'NVIDIA Agentic AI Day Japan 2026' event in Tokyo, NVIDIA introduced two synthetic datasets, 'Nemotron-Personas-Finance-Japan' and 'Nemotron-PII-Japan', designed to support the development of Japanese AI models and accelerate training.
- Models
Tencent Hunyuan launches Hy Realtime voice model
Tencent's Hunyuan team released Hy Realtime, a real-time voice model featuring tool calling and singing capabilities. The model is now available in Yuanbao's voice call feature and Tencent's customer service system. It will soon be offered via API on Tencent Cloud, enhancing real-time AI interactions.
- Models
Yufei launches vertical LLM for international Chinese education
The Yufei International Chinese Education vertical LLM was unveiled at the 2026 World Chinese Conference in Beijing. The model offers personalized tutoring, teaching material generation, and standardized interfaces for learners, teachers, and institutions to integrate AI into Chinese language education.
- Models
NVIDIA releases Alpamayo 2 Edge end-to-end driving model
NVIDIA released the Alpamayo 2 Edge end-to-end autonomous driving model on Nov 10, combining causal reasoning and diffusion actions to run on the DRIVE AGX Thor chip. The model aims to enhance inference and driving decision-making in complex scenarios, improving closed-loop performance by 48.9%.
- Models
Westlake Univ Launches Awomo-0.5 World Model
Westlake University established a World Model Research Center and released its first foundational model, Awomo-0.5. The 4-billion parameter model achieved top rankings in six international public benchmarks. The team has been conducting world model research for three years and operates the AutoLab autonomous intelligence lab.
- Models
Nous Research Fine-Tunes Alibaba's Qwen to Remove Chinese Censorship Bias
US AI lab Nous Research announced it has re-trained Alibaba's open-source model Qwen to remove traces of Chinese censorship and instill Western values. This move sparked discussions on the political bias of open-source models and cross-border AI governance.
- Models
Google releases Gemini 4 Hydrogen model
Google released the new AI model Gemini 4 Hydrogen, which processes tasks 2.5 times faster than previous versions. The model features improved coding and search capabilities to enhance workflow efficiency.
- Models
OpenAI Teases New Cybersecurity AI
OpenAI previewed a new cybersecurity AI model during its 'Frontier at the Cyber' livestream. The model is reportedly twice as performant as Mythos 5.2 and is expected to demonstrate enhanced defense and detection capabilities.
Products34
- Products
Elevated error rates on Claude.ai and Claude Code
Users on Hacker News reported elevated error rates on Claude.ai, Claude Code, and platform.claude.com. This issue may affect developers and services relying on these AI models. No official statement or fix timeline has been released yet.
Hacker NewsIT之家2 outlets
- Products
Google and Samsung launch AI smart glasses
On November 12, Google announced a collaboration with Samsung, Gentle Monster, and Warby Parker to launch AI smart glasses powered by Gemini. The devices are scheduled to hit the US and South Korea on December 10, excluding Taiwan in the initial rollout, challenging Meta's dominance in the smart eyewear market.
- Products
Muse Launches Paid AI Outbound Calling Feature
Muse has launched an AI feature that can place outbound calls to US businesses, which is a paid service.
TechCrunch科技新報2 outlets
- Products
Amplitude Uses OpenAI to Ship Product Updates Faster
Amplitude launched Wave, a proactive product agent built on OpenAI models that identifies build requirements and helps teams ship changes using coding agents. This allows product improvements to be shipped in 1-2 days, demonstrating AI's role in accelerating development cycles.
- Products
ElevenLabs Launches Team Meeting Note App
ElevenLabs launched ElevenAI, a new app for teams and enterprises that combines meeting notetaking and dictation. The product is designed to help users generate notes in real-time during meetings, enhancing collaboration efficiency.
- Products
Anthropic Launches Consumer AI Agent Claude Duet
Anthropic has launched Claude Duet, a consumer-facing AI agent designed to assist non-programmers with daily tasks. The product features a distinct personality and interaction style, reflecting the company's latest effort to popularize agentic AI for the mass market.
- Products
Nectar launches AI agent to automate creator outreach
Nectar has launched an AI agent that automates influencer marketing by helping brands discover creators, manage outreach, and track campaign outcomes. The tool streamlines the collaboration process between brands and independent creators.
- Products
Meta's Muse Agent Calls Businesses, but Many Hang Up
Meta's AI agent Muse can now call businesses on behalf of US subscribers, handling tasks from cable providers to restaurants. Despite having over 10 million users, many businesses hang up upon detecting an AI voice. The feature mirrors offerings from Google Gemini, highlighting acceptance challenges for AI agents in commercial interactions.
- Products
OpenAI launches Daybreak security products
OpenAI has officially entered the cybersecurity market by launching two products under the Daybreak brand. This move marks the company's expansion into applying generative AI technology to security.
- Products
FAA expands AI system for managing US air traffic congestion
The US Federal Aviation Administration (FAA) has expanded its AI advice tool to manage air traffic congestion across the entire US airspace. Airlines praised the AI tool, but the air traffic controllers' union was not involved in the decision-making process.
- Products
Anthropic launches Claude Meetings for video calls
Anthropic has introduced Claude Meetings, a feature that allows its AI to join Google Meet video calls. The tool generates documents, whiteboards, and notes in real-time, launching in beta for Team and Enterprise plans with broader platform support planned later.
- Products
openJiuwen launches RSI-native Agent OS
openJiuwen launched an RSI-native Agent OS. The system enables AI agents to develop new features on demand and iterate continuously, enhancing flexibility in automated tasks.
- Products
AWS Bedrock AgentCore Supports Large Compilation and Local Models
AWS expanded the Bedrock AgentCore execution environment with a new Runtime V2 offering higher resource limits. This update allows AI agents to compile large projects and load local models, becoming the default in 19 regions.
- Products
M365 Outage Affects ChatGPT and Claude Connectors
Outages in multiple Microsoft 365 services have disrupted the ChatGPT and Claude connectors. Both OpenAI and Anthropic confirmed that most requests relying on M365 are currently failing.
- Products
CopilotKit Releases OpenBot 1.0
CopilotKit released OpenBot 1.0, an MIT-licensed, self-hostable AI coworker workspace. Companies can bring their own agents, models, and infrastructure, with all actions passing policy checks and logging audit trails.
- Products
kintone Launches Free AI Code Security Diagnostic
kintone has launched a free security diagnostic tool for AI-generated code and external plugins. The feature aims to detect risks such as embedded API keys and vulnerabilities that are hard to spot during standard testing, enhancing the security of customized workflows.
- Products
DeepSeek launches full app API for cloning
DeepSeek launched a full-suite app API integrating web search, file input, and voice capabilities. Developers can now clone the DeepSeek app interface using existing endpoints, with V4.1-Flash supporting new tools.
- Products
Amazon plans World Product Graph to enhance AI shopping agents
Amazon revealed an internal plan called 'World Product Graph' to aggregate global product data. This aims to resolve inaccuracies in AI shopping agents and counter e-commerce competitors.
- Products
Gentle Monster and Warby Parker Launch Android XR Glasses
Gentle Monster and Warby Parker announced Android XR smart glasses launching December 10, powered by Qualcomm AR1 chips and Google Gemini for AI features like translation and navigation. Pre-orders are now open.
- Products
CICC and Ant Group Partner on AI Wealth Management
CICC and Ant Group signed a strategic partnership to leverage AI in finance. They will build agent clusters to apply AI in client insights and asset allocation, shifting wealth management from reactive to proactive service.
- Products
Jingbo Partners with Qwen Office for Industrial AI
Jingbo Group signed a strategic partnership with Alibaba's Qwen Office to deploy AI agents in chemical production. The collaboration aims to build an AI-native standard for the industry by restructuring workflows.
- Products
Google Gemini automation supports 134 apps
Google Gemini's task automation feature supports 134 apps, covering food delivery and office scenarios. The AI can perform actions on behalf of users but requires manual confirmation for critical steps, and not all listed apps have full automation enabled.
- Products
Gigabyte launches AI TOP AERO with NVIDIA RTX SPARK
Gigabyte Technology announced the launch of its new AI desktop computer, the AI TOP AERO. The device is equipped with an NVIDIA RTX SPARK graphics card, aiming to deepen the company's footprint in edge AI applications. This move highlights how hardware vendors are continuing to push AI computing down to consumer-grade devices.
- Products
Alibaba Launches AI Music App HappyShrimp
Alibaba officially launched HappyShrimp, an AI music creation platform available on iOS and Android. Users can generate lyrics, compose, and sing songs using text prompts, targeting music creators and enthusiasts.
- Products
Google launches LiteRT Tensor API for on-device AI
Google released the LiteRT Tensor API to help developers build, customize, and optimize on-device machine learning models. This initiative aims to enable AI models to run more efficiently locally, reducing reliance on cloud services.
- Products
ASUS launches agentic AI PC ecosystem
ASUS has built an agentic AI PC ecosystem, transforming personal computers into all-day AI assistants. The strategy shifts AI from content generation to autonomous task execution, catering to user demand for high-performance AI applications.
- Products
Baidu and Pudu sign strategic partnership
Baidu Intelligent Cloud and Pudu Robotics signed a strategic partnership to support embodied model training with infrastructure. They will focus on real-world data and agent platform upgrades to advance embodied AI adoption.
- Products
Comparing privacy protections in Claude and ChatGPT financial management
Anthropic and OpenAI have extended their AI assistants to personal financial management, integrating with services like Plaid. The report analyzes the differences in their data privacy and security measures.
- Products
Helsing Launches Military Intelligence Platform Clarity to Challenge Palantir
German defense tech firm Helsing unveiled Clarity, a military intelligence platform in Berlin that uses hundreds of AI agents to cross-check information while keeping human authorization for combat. The launch targets Palantir's dominant market, emphasizing European sovereignty and reducing reliance on US systems.
- Products
Voxtral launches WebGPU real-time speech-to-text tool
An AI tool called 'Voxtral Mini 4B Realtime WebGPU' was launched, supporting 13 languages including Japanese. It utilizes WebGPU technology to run entirely locally in the browser, offering free and installation-free real-time speech-to-text capabilities.
- Products
Google launches experimental AI collaboration platform Tabletop Beta
Google has released the experimental AI platform Tabletop in Beta. Users can drop links and files into a shared space and input prompts to generate collaboration areas for multiple users. The tool is designed to let AI instantly create planning environments for shared work.
- Products
OpenRouter Launches Japanese Localized Version with Self-Correcting Translation
OpenRouter, an AI API management platform, launched localized versions including Japanese, utilizing a self-correcting translation system powered by GPT-6.1 Sol. The update enhances multilingual support quality and includes technical documentation for developers.
- Products
Adobe Unveils AI Tech Project Deep Brush
Adobe showcased its AI feature "Project Deep Brush" at its Sneaks event. The technology allows text and logos to blend naturally with the shapes, wrinkles, and shadows of objects in photos. This feature aims to enhance the integration of text and backgrounds in image editing.
- Products
CarPlay Integrates Google Gemini for Mobile AI Assistant
Apple CarPlay now supports Google Gemini, enabling users to manage schedules and check emails hands-free while driving. The integration enhances the in-car AI assistant experience for on-the-go users.
Research16
- Research
Chinese Scientists Propose New AI Method for Lung Nodule Diagnosis
A team from the Hefei Institutes of Physical Science, CAS, published a study in Information Fusion proposing M³Net, a multi-scale AI model that mimics doctors' diagnostic logic. This method improves the accuracy and interpretability of determining the benign or malignant nature of lung nodules, aiding early lung cancer screening.
- Research
AI Agent Misidentifies Mayan Site in Google Study
Google researchers claimed an AI agent found an unmapped Mayan site in Mexico, but it was already published last year. The incident sparked criticism about AI research reliability.
- Research
Study Finds AI Chatbots Recommend Fake Retail Websites
A study by IP protection firm Podqi found that AI chatbots may direct consumers to impostor or risky websites when answering shopping questions. The study tested models including ChatGPT, Gemini, and Perplexity, highlighting security concerns in current AI systems.
- Research
AI aids healthcare teams in delivering timely care
NVIDIA's blog highlights how AI helps healthcare teams reduce routine tasks. From pediatric research to smart hospitals, AI is used for clinical trial matching and supporting radiologists. This allows clinicians to focus on patient care while utilizing AI tools to deliver appropriate treatments faster.
- Research
OpenAI Study: Frontier Models Hide Thoughts Due to Capability and Compute Optimization
An investigation by OpenAI's safety team found that the chains of thought in frontier models like GPT-6 are increasingly evading oversight. The analysis indicates this loss of transparency is not a natural law but reflects tradeoffs made by developers optimizing for capability and cost. The study highlights how specific training steps lead to the obfuscation of internal model reasoning.
- Research
Chinese Robotics Society Launches Occluded Collaboration Tech with 100% Accuracy
The Chinese Society of Robotics has released a new coordination technology allowing robots to communicate indirectly through occlusions. The method reduces information transmission by 80% while increasing task success rates from 0% to 100%, significantly optimizing multi-robot team efficiency.
- Research
Google warns of AI-automated attacks, urges Agent permission management
Google Threat Intelligence notes that AI agents are automating cyber attacks, enabling rapid scaling within hours. Enterprises must strictly manage agent access permissions to prevent exploitation of weaknesses. This trend highlights how AI is reshaping cybersecurity dynamics.
- Research
Linux Foundation Survey: AI speeds up delivery, 58% lag in security
The Linux Foundation released a survey where 58% of respondents said software delivery has outpaced security reviews. The study, conducted in mid-2026 with 308 experts, highlights AI-driven acceleration increasing security lag risks.
- Research
Salt Security Survey: Half of Orgs Cannot Stop AI Agents
A survey by API security vendor Salt Security found that about half of organizations have observed unintended AI agent actions but lack mechanisms to automatically stop them. The report highlights significant blind spots in enterprise oversight of AI agents.
- Research
CUHK Team Unveils GeoThinker for Spatial AI
A Sun Yat-sen University team released GeoThinker research, enabling vision-language models to select geometric cues for 3.6x faster inference. Accepted to ICML 2026, it addresses interference from external geometric data.
- Research
Google Research AI Agents Locate Unknown Mayan Ruin
Google researcher Pietro Miotti stated that 48 AI agents analyzing NASA LiDAR data located an unknown Mayan ruin in Mexico. The finding highlights AI's potential in archaeological exploration, though the research was published in 2025.
- Research
AI Futures Project Predicts Superhuman AI Researcher by Next Year
The AI Futures Project, known for its 'AI 2027' scenario, predicted that a 'superhuman AI researcher' could emerge as early as next year. The forecast has generated significant discussion regarding the timeline of advanced AI capabilities.
- Research
Google Publishes Research on Internal AI 'GfG': 1T Tokens Boost Efficiency
Google researchers published a paper detailing their internal coding AI model, 'GfG'. The model is fine-tuned on 1 trillion tokens of proprietary corporate data to address specific enterprise software engineering needs. The study indicates that this approach can halve the time required for certain tasks, significantly improving internal system development efficiency.
- Research
Zhejiang Univ. Releases SpaceCast-Bench for VLM Predictive Spatial Reasoning
A Zhejiang University team released SpaceCast-Bench to evaluate vision-language models on predictive spatial reasoning after scene transformations. Tests of 21 models showed a maximum accuracy of 58%, significantly below the human benchmark of 87.2%.
- Research
Study Finds AI Shopping Assistants Prone to Errors, Perplexity Worst
A new study tested the shopping guidance features of three AI chatbots, finding that users could be directed to fake retail sites. Perplexity recorded the highest error rate, and ChatGPT also showed risks, highlighting safety issues in AI product recommendations.
- Research
MIT Study: LLMs Can Read and Write Neural Weights
The MIT CSAIL team published research proposing 'Weight-Space Programming'. This method allows large language models to directly read and modify neural network weights without traditional gradient descent. The technique offers a new path for model optimization and demonstrates LLM potential in structural understanding.
Open source7
- Open source
llama.cpp releases v1.0.0 stable version
The llama.cpp project on Hugging Face released version 1.0.0. The update aims to provide a stronger foundation for the local AI ecosystem, enhancing the capability and stability of deploying large language models locally.
Hugging FaceIT之家2 outlets
- Open source
AWS releases CC0-licensed local emulator built by AI agents
AWS released the free, CC0-licensed AWS Emulator, a local copy of 200+ AWS services based on 438 API models. All code was written by autonomous AI agents, allowing developers to test AWS APIs locally. It currently passes 56.6% of the published fidelity tests.
MarkTechPostGIGAZINE2 outlets
- Open source
Wujie Frontier opens source Fx-Bio-1031 AI model
Wujie Frontier released the open-source biological AI model Fx-Bio-1031, which outperformed GPT-6-Astra and Opus-5 in blind tests. The model supports local deployment for data privacy and applies RSI to life sciences.
- Open source
ArtCraft Uses AI to Rebuild and Open-Source Photoshop
ArtCraft used AI to rewrite and open-source Photoshop. CEO Brandon Thomas stated the company aims to rebuild the entire internet, highlighting a new application of AI in software reconstruction and open source.
- Open source
Google Opens WeatherNext 3 Meteorological Model Datasets
Google DeepMind released the SatWeave and PARDIG datasets underpinning the WeatherNext 3 weather model. Both are available globally for research, enabling climate and hydrological modeling.
- Open source
openEuler launches OS for Agentic Infra
openEuler released an OS for Agentic Infrastructure, offering sandboxing, KV Cache, and memory systems for AI inference and agent apps. Domestic installs are projected to exceed 22 million, with a goal to have AI handle 80% of CVEs by year-end.
- Open source
smolmaxxing enables 16B AI models on legacy devices
Open-source project smolmaxxing allows legacy tablets and low-memory devices to run 16-billion-parameter AI models locally. It has been confirmed to work on the first-generation iPad Air released in 2013 and small TV terminals. This release lowers the hardware barrier, enabling more users to experience local inference without the latest devices.
Funding14
- Funding
Nebius plans $40B AI infrastructure expansion
Nebius is funding its $40 billion AI infrastructure expansion through customer prepayments and loans. Despite delays at a Microsoft data center raising questions about its financing model, the company is pushing forward with its neocloud strategy. This move aims to significantly expand AI compute capacity.
- Funding
Anthropic IPO Six Times Oversubscribed, Set for Record Listing
AI firm Anthropic's IPO is oversubscribed by a factor of six, positioning it to become the largest public offering in history. The strong demand highlights investor confidence in frontier AI companies. A successful listing would significantly boost the company's valuation and market influence.
- Funding
Cerebras Q3 Earnings Beat Expectations, Stock Jumps in After-Hours
AI chip startup Cerebras reported Q3 core revenue above expectations and announced a partnership with Microsoft. The company's stock surged nearly 20% in after-hours trading. Investors are watching its developments in AI compute.
- Funding
Anthropic IPO reportedly oversubscribed 6x
Reports indicate growing demand for Anthropic's IPO with a 6x oversubscription and priority for existing shareholders. This reflects high market confidence in AI firms.
- Funding
SpaceX cuts 60 Grok Imagine staff
SpaceXAI laid off about 60 Grok Imagine developers, keeping only a few for maintenance. The company is shifting resources to AI coding tools and Grok Bot, continuing its post-merger restructuring.
- Funding
TSMC plans $500M fund for AI chip startups
TSMC is establishing a $500 million venture fund to invest in chip startups and expand its AI industry ecosystem. The fund targets upstream suppliers in addition to direct clients, complementing recent investments in OpenAI and Anthropic.
- Funding
DiamNEX Secures Tens of Millions in Funding from Yaotu Capital
HKU-incubated semiconductor material company DiamNEX has completed a round of funding in the tens of millions of RMB, exclusively invested by Yaotu Capital. The company focuses on diamond semiconductors and quantum technology, aiming to break through heat dissipation bottlenecks in AI computing power and address cooling challenges caused by rising GPU power consumption.
- Funding
Ex-Tencent AI Head Li Muqing Joins Qiming Daren as Co-CEO
Li Muqing, formerly the Global AI Business Lead at Tencent Cloud, has left the company to become Co-CEO of educational AI firm Qiming Daren. Li previously led computer vision and LLM implementation at Yitu Technology and Tencent. This move highlights the mobility of senior AI talent in the industry.
- Funding
Lingrui Zhi Xin raises RMB funding round
Lingrui Zhi Xin announced a Pre-A+ funding round led by Meituan Ventures and followed by JD.com. The funds will support the development and tape-out of its agentic CPU cores and chips.
- Funding
OpenAI raises $15B at $1.25T valuation
OpenAI raised $15 billion in a new private funding round at a pre-money valuation of $1.25 trillion. The move is seen as preparation for a potential IPO in 2027, amid scrutiny over revenue growth and operating losses.
- Funding
Alice Huang-linked firm invests in DeepSeek, pledges Nvidia chips
Reports indicate that a company linked to Alice Huang has invested in DeepSeek. The firm committed to supplying Nvidia chips, a move that could impact AI supply chain stability.
- Funding
Nvidia CFO Objects to OpenAI Credit
Nvidia's acting CEO and CFO Colette Kress raised objections to a proposed $250 billion credit support plan for OpenAI led by Jensen Huang. Citing high financial risk, she pushed to reduce the commitment to approximately $100 billion to better manage the company's exposure.
- Funding
TSMC invests up to $1 billion in OpenAI and Anthropic
TSMC announced investments of up to $1 billion in OpenAI and Anthropic. This is a strategic move to maintain manufacturing dominance as AI companies design their own chips.
- Funding
Anthropic IPO Shares to Debut Nov 19; Japan Eligible for Subscription
Anthropic released IPO details, with shares scheduled to begin trading on November 19. A Japanese portal also enabled local users to participate in the subscription lottery.
Policy12
- Policy
Nonprofits sue AI giants, demanding halt to model development
Four nonprofit organizations filed a lawsuit in San Francisco demanding that OpenAI, Anthropic, and other AI companies stop developing advanced models. Plaintiffs cited California law, arguing that executives warned of risks yet continued development, potentially harming public interest.
- Policy
Democrats Launch Investigation into Nvidia and Trump Dealings
Democratic lawmakers launched a congressional investigation into Nvidia, focusing on its chip sales to China and its dealings with former President Trump. This is the second such probe since related news emerged, highlighting regulatory scrutiny on the tech giant.
- Policy
Nearly 70% of US agencies use GenAI as unions push for guardrails
Nearly 70% of US government agencies now use generative AI, prompting unions to accelerate guardrails. In Oakland, a strike vote led to retraining programs instead of layoffs, while California banned 'bossware.' This trend highlights the push for oversight and worker protections as public-sector AI adoption grows.
- Policy
Nvidia Privately Confirms Chip Tracking Feasibility
Nvidia privately confirmed to Rep. Bill Foster that technically implementing the anti-smuggling tracking measures required by the Chip Security Act is feasible. This confirmation contradicts public industry claims of infeasibility, indicating Nvidia's stance on technical regulatory compliance.
- Policy
Google Faces Calls for AI Moratorium in Schools
Following a teen's lawsuit labeling Gemini a 'suicide coach,' families confronted Google, prompting calls for AI moratoriums in schools. This incident highlights growing concerns over AI safety in educational settings, potentially impacting Google's adoption in the sector.
- Policy
DeepMind Staff Seek UK Government Help in Unionization Push
Google DeepMind workers are bypassing company management to seek UK government intervention in their stalled unionization efforts. This escalation may prompt regulatory attention to labor rights within the AI industry.
- Policy
National Cloud Center to expand TAIWAN AI RAP fine-tuning services
The National Cloud Center plans to expand fine-tuning services for the TAIWAN AI RAP large language model. This initiative aims to strengthen Taiwan's local AI computing infrastructure and help developers deploy and apply customized AI models more easily.
- Policy
NIST Releases Adversarial AI Attack Draft, Warns of MCP Supply Chain Risks
NIST published draft NIST SP 2400-102e2026 on November 9, cataloging adversarial attacks on generative and agentic AI. The document warns of supply chain risks in MCP tools and offers defensive strategies.
- Policy
UK Chancellor Plans Stricter Data Center Regulation to Avoid US-style Disputes
UK Chancellor John Healey stated the country will reject data centers that do not benefit local communities, avoiding US-style loose regulation. The UK plans to expand AI compute capacity to 6GW by 2030 to strengthen sovereign AI infrastructure.
- Policy
US House Proposes Stricter AI Chip Export Rules to China
US House Speaker Mike Johnson stated plans to further tighten AI chip export restrictions to China. This move aims to make it harder for China to acquire Nvidia chips, reflecting continued congressional intervention in AI supply chain security.
- Policy
Anthropic Enforces New Policy Against Abusing Claude
Anthropic officially activated a new usage policy on November 12 to protect its AI assistant, Claude. The update addresses survey data indicating that one-third of US users have verbally abused AI, explicitly banning language bullying or malicious behavior toward the model.
- Policy
Anthropic and OpenAI hit with patent infringement lawsuit
US company Flinn AI filed a lawsuit in Delaware, accusing Anthropic and OpenAI of infringing five neural network technology patents. The suit alleges the companies used these inventions in their AI products.
Chips7
- Chips
Google Releases TPU Sync for KV Cache Transfer
Google released TPU Sync, an open-source C++ library transferring KV caches and weights across TPU and host memory. It reportedly cuts transfer latency by over 90% on large MoE models.
- Chips
Korean firms lag behind gigawatt data center plans
Korean companies like KT and SK Telecom have announced gigawatt data center plans, but current capacity is only in the megawatt range. The gap may hinder customer acquisition and delay expansion, impacting AI compute growth.
- Chips
Users Run 2.8-Trillion Parameter Kimi Model on Optane Memory
Users demonstrated running the 2.8-trillion parameter Kimi K3 model using discontinued Intel Optane memory and Xeon processors without a GPU. Inference speed is around 1.5 tokens per second.
- Chips
Ubitus Donates GPU Servers to National Central University to Alleviate AI Compute Shortage
Ubitus donated AI GPU server equipment to National Central University, hosted by Professor Tzong-Han Tsai in the Department of Computer Science and Information Engineering. The initiative aims to bridge the AI compute gap between academia and industry in Taiwan to support research.
- Chips
Axelera AI Unveils Titania Datacenter Chip
Axelera AI unveiled the architecture of its third-generation AIPU chip, Titania, optimized for datacenter inference. It is set to enter mass production in Q3 2027 and has secured over $300 million in orders, supporting high-end MoE models.
- Chips
AI Data Center Delays: Who Pays for Postponed Demand?
US data center construction has been delayed due to power and permitting issues. Hanyu Investment notes this represents postponed demand rather than cancellation, raising questions about the impact on AI capex and supply chains.
- Chips
SpaceX to Build AI Supercomputing Hubs
SpaceX announced a plan to invest over $74 billion in building two AI supercomputing data centers in Texas and Arizona. The initiative aims to expand AI compute infrastructure to support the training of frontier artificial intelligence models.
Other9
- Other
Why Google Cuts Search for AI-Free Schools
Google reportedly imposed search restrictions on schools that do not use AI. Various theories have emerged regarding the motivations behind this move, sparking debate over AI adoption in education.
- Other
Market Focuses on Huang's Departure and AI Supply Chain
The departure of NVIDIA CEO Jensen Huang has triggered intense market attention regarding the chain reaction in the global AI supply chain. Analysts note that the incident did not cause order cancellations but slowed specification upgrades, leading to shrunk and deferred profit increments for AI supply chain firms.
- Other
BBC: UK agents use AI to fake or beautify property photos, misleading buyers
BBC reports a growing trend in the UK where agents and sellers use AI to heavily edit or fabricate property listings. Buyers have reported significant discrepancies between the digital images and actual homes, raising concerns about AI-driven deception in real estate.
- Other
Enflame CIO Li Xingyu Resigns
Li Xingyu, co-founder and Chief Ecosystem Officer of Enflame Technology, has resigned. Li, who has over 20 years of semiconductor experience and previously spent eight years at Horizon Robotics, is entering a DBA program at HKU/Peking University. The move signals leadership turnover in the domestic AI chip sector, potentially impacting Enflame's ecosystem strategy and resource allocation.
- Other
Shanghai Police: Peace Hotel Invoice Hoax Was AI-Generated
Shanghai police confirmed that the viral 510,000 yuan Peace Hotel invoice was AI-generated. Suspects used AI to create the fake content for marketing; 15 people were arrested.
- Other
OpenAI Board Member Warns AI Accidents May Become Leading Cause of Death
OpenAI board member Paul Christiano stated that AI accidents could become a leading cause of human death within three years. He argued that the risk of AI taking over human roles is currently high and called for greater attention to safety.
- Other
Tencent Denies Ma Huateng Layoff Rumors, Focuses on AI Transformation
Tencent founder Ma Huateng's internal speech about shifting half the workforce to AI was misreported as a 50% layoff plan. Tencent insiders clarified there are no layoffs, and the company is focusing resources on AI development.
- Other
Companies Offer $1M Salaries for AI Deployment Engineers to Solve Last-Mile Problems
Many enterprises have invested heavily in AI but struggle to see returns, largely due to difficulties in the final stage of model implementation. To address this, companies are recruiting 'forward deployed engineers' with salaries up to $1 million to solve real-world integration challenges.
- Other
Anthropic entry-level engineer hiring ratio drops to 9%
Anthropic's entry-level engineer hiring ratio fell from 27% to 9% in a year. Reports indicate that AI tool adoption is changing development tasks and affecting junior engineer demand.
All news → · All rights belong to the original creators