AI news for Friday, 6 November 2026
Stories 61 to 120 of 135, most important first. Back to the first page →
-
Study finds AI agents waste 18% of tokens on unnecessary tasks
A new US academic study quantifies how AI agents waste tokens and money during agentic workflows. Researchers tracked token usage across six models—Kimi-K3, Gemini 3.1 Pro, GPT-5.6 Terra, GPT-6 Astra, Claude Sonnet 5,…
1 sourceUnite.AI -
OpenEvidence integrates ACR patient fact sheets into clinical AI workflows
OpenEvidence and the American College of Rheumatology (ACR) announced a collaboration on November 6, 2026, to embed ACR’s patient education fact sheets into OpenEvidence’s medical AI platform. The English- and…
1 sourceUnite.AI -
Luxury Brands lean into handcrafted art, artists say
Prada released a 2026 holiday campaign featuring hand‑drawn animation created with Disney animator John Pomeroy. The teaser shows pencil sketches coming to life, highlighting the skill of analog drawing. The campaign…
1 sourcenbcnews.com -
Researchers autoformalize linear regularization theory in Lean with AI agents
A team of researchers has developed AutoIP, a Lean library that uses AI agents to autoformalize classical linear regularization theory and its foundational functional analysis. The project leverages AI to generate…
2 sources primary sourcearXiv cs.AI -
BlockLock fixes LLM inference batching for reproducible outputs
BlockLock is a new method for ensuring that large language model (LLM) inference produces the same output for a given request, no matter what other requests are in its batch. Existing approaches either constrain every…
1 source primary sourcearXiv cs.AI -
Nvidia open source AI Aerial and Sionna tools accelerate 6G research
Nvidia has expanded its open‑source wireless suite, adding AI Aerial and Sionna to help researchers prototype AI‑native 6G networks. Over the past year global adoption of the tools grew tenfold, downloads passed one…
1 source primary sourceNVIDIA Blog -
Airbnb discloses option to invest in Brian Chesky's stealth AI lab
Airbnb has disclosed in a filing that it holds an option to invest in and license technology from a stealth AI startup controlled by its founder and CEO, Brian Chesky. The disclosure sheds light on the previously…
2 sourcesUnite.AItheinformation.com -
Aws builds Serverless MCP server on ECS Fargate for AI agent data lake access
aws introduced a serverless model context protocol (mcp) server built on ecs fargate to give ai agents a single gateway to enterprise data lakes. the system removes the need for per‑dataset integration code by using a…
1 source primary sourceAWS Machine Learning Blog -
Amazon adds remote Claude Code support on Bedrock AgentCore with security controls
Amazon’s Machine Learning Blog details a new pattern that lets developers run Claude Code as a remote agent on Amazon Bedrock AgentCore. Each coding task is executed in its own isolated microVM inside the user’s AWS…
1 source primary sourceAWS Machine Learning Blog -
Orca Security’s head builds partner platform with AI in a weekend
Shannon McWilliams, head of global distribution at Orca Security, used AI tools to prototype the company’s Partner Success Platform over a weekend. The platform now includes AI-driven features like deal coaching and…
1 sourceUnite.AI -
Amazon Quick defines four agentic primitives based on When, What, Who, and How
Amazon Quick introduces a taxonomy that breaks down agentic work into four plain questions: When, What, Who, and How. The framework groups the building blocks into triggers that start work, agents that own the…
1 source primary sourceAWS Machine Learning Blog -
Amazon SageMaker Inference supports GPU partitioning
Amazon SageMaker Inference now supports GPU partitioning via NVIDIA Multi‑Instance GPU (MIG), allowing users to split a single GPU into multiple isolated partitions. This lets more models run on a single accelerator,…
1 source primary sourceAWS Machine Learning Blog -
Taxbit launches role-aware AI assistant on Amazon Bedrock AgentCore
Taxbit, a digital-asset tax compliance platform, has deployed an AI assistant called Taxbit Assist within its product dashboard. Built on Amazon Bedrock AgentCore, the tool is designed to help compliance teams at…
1 source primary sourceAWS Machine Learning Blog -
iQIYI reconstructs lost 1938 sci-fi film with NadouPro AI tools
iQIYI’s NadouPro platform helped an eight-person team reconstruct Visiting Shanghai after Sixty Years, a lost 1938 Chinese sci-fi comedy, using only still photos, posters, and newspaper clippings. The project,…
1 sourceUnite.AI -
Rising rates may strain AI firms’ debt and growth, analysts warn
Analysts warn that higher interest rates from the US Federal Reserve and other central banks could intensify pressure on AI companies, which have borrowed heavily to fund expansion. Charles Lagares, COO of EBC…
1 sourcehotminute.co.uk -
CBD picks Harvey AI for legal research, drafting and document review
Commercial Bank of Dubai (CBD) has chosen Harvey AI to power its in-house legal team’s workflows, covering capital markets, investment, regulatory and commercial matters. The platform integrates with Microsoft Word and…
1 sourceUnite.AI -
Liontrust highlights NVIDIA’s AI infrastructure role in Q3 2026 letter
Liontrust’s GF Global Innovation Fund Q3 2026 letter notes NVIDIA as a top holding, up 14% in the quarter. The fund rebuilt its position in July after semiconductor stocks lost about $3 trillion in market cap amid…
3 sourcesfinance.yahoo.comfool.comnote.com -
Nvidia in talks to guarantee financing for Nscale data center serving Anthropic
Nvidia is discussing a guarantee for construction financing of a West Virginia data center that neocloud Nscale is building for Anthropic, The Information reported. If Nscale defaults, Nvidia could become responsible…
1 sourcebusinessinsider.com -
AI companies may buy work files and emails, Wired reports
AI firms are reportedly buying professional documents, emails and other workflow data to train models. The industry, led by startups like Mercor, Micro1 and Handshake, is buying graduate‑student papers, corporate…
1 sourceWired AI -
Microsoft reduces reliance on CoreWeave for AI compute, cutting its stock by $3
Microsoft is reportedly shifting AI training workloads in-house, reducing its dependency on CoreWeave, a major GPU provider. The company signed a multi-year deal with CoreWeave in 2023, but the shift has caused…
1 sourcecryptobriefing.com -
The Decoder profiles 27 neolabs betting on new AI research paths
The Decoder’s latest Frontier Radar issue examines the rise of "neolabs," independent AI companies founded by former researchers from major tech giants. Despite OpenAI and Nvidia declaring the arrival of the AGI era,…
1 sourceThe Decoder -
Cashfree adds WebMCP tools to Hosted Checkout for AI agents
Cashfree Payments announced on November 6, 2026, that its Hosted Checkout now supports WebMCP, enabling AI agents to automate parts of the payment process in browser-based checkouts. The tools allow agents to read…
1 sourceUnite.AI -
AI decompilation may speed up GTA VI PC port, former Rockstar tech director says
A former Rockstar tech director suggests AI-assisted reverse-engineering of console games could force the studio to accelerate its PC port of Grand Theft Auto VI. Obbe Vermeij, who left Rockstar in 2009, recalled past…
1 sourceArs Technica AI -
OpenAI tests demand response in Georgia hub with 1,000MW cut
OpenAI launched the Grid Protectors Program on November 2, 2026, to reduce electricity demand from its computing hubs during grid strain. The initiative involves contracts with utilities and regulators to adjust power…
1 sourcenote.com -
Cramer warns neocloud earnings may test AI data center stocks
CNBC's Jim Cramer expressed skepticism about the recent market rebound, describing Friday's gains as a "dead cat bounce" following a turbulent election week. He noted that investors remain wary of the artificial…
1 sourceCNBC Technology -
Researchers introduce efxplorer for Cooperative Multi-Agent reinforcement learning
A team of researchers has proposed EFXPLORER, a new framework for multi-agent reinforcement learning (MARL). The method aims to address the challenge of achieving cooperative behavior by balancing exploration and task…
1 source primary sourcearXiv cs.AI -
AI is dating dealbreaker in tech hub
In San Francisco, where artificial intelligence (AI) is booming, dating apps reveal that AI has become a significant factor. Users often include statements like 'We’ll get along if you don’t use AI' or 'You should not…
1 sourceThe Verge AI -
Anthropic adds enterprise security checks to Claude with inference hooks
Anthropic introduced new security and management features for Claude Enterprise on November 5, 2026. The key addition is Inference hooks, a system that lets companies integrate their own rules to block sensitive data…
1 sourcenote.com -
Russia and US move to block binding autonomous weapons rules at UN conference
Plans to fill a gap in global governance of autonomous weapons are in jeopardy as a bloc led by Russia and the United States is using a UN conference this month to block proposals for a legally binding set of rules.…
1 sourceThe Guardian AI -
Agility Robotics releases room-scale cable-driven home robot for $795
Agility Robotics unveiled a novel home robot called Stringman, designed for room-scale tasks like tidying. The device uses cable-driven mechanics to move around and interact with objects, priced at $795. The company…
1 sourceIEEE Spectrum Robotics -
Zapier Sidekick automates deal reports, Slack channels and more for users
Zapier Sidekick automates repetitive tasks for its own teams and customers. Five real-world examples show how it handles deal breakdowns, out-of-office replies, calendar scheduling, AI feature development, and Slack…
1 sourceZapier Blog -
Tesla unveils AI roadmap, adopts 'super intelligence' branding
Tesla has unveiled a new AI accelerator roadmap that includes the Dojo D3 training accelerator, which will use stacked HBM memory instead of SRAM-centric architecture. The company also adopted the term 'super…
1 sourceTom's Hardware -
Zapier finds Claude Code uses Markdown mirrors and hidden Haiku 4.5 model
Zapier built a tool to test how AI agents like Claude Code interpret its integration pages. By comparing production and preview versions of the Slack integration page, they discovered Claude’s WebFetch tool first…
1 sourceZapier Blog -
RunSybil finds CPython flaw enabling code execution via chat templates
Cybersecurity firm RunSybil discovered a zero-day vulnerability in the CPython interpreter that allows attackers to escape Python sandboxes, such as Jinja chat templates. By crafting a malicious model file—demonstrated…
1 sourcerunsybil.com -
The Rise of the Executive AI Briefing: How CEOs Start Their Day in the Agentic Era
These agents prioritize changes since yesterday—revenue shifts, pipeline risks, operational anomalies—and allow executives to ask follow-up questions in natural language. Unlike traditional dashboards, they rely on a…
1 sourceUnite.AI -
Trump urges states to welcome AI data centers after ballot defeats
President Trump said Friday he will work with states that want AI data centers, calling them "GOLD" and saying only "foolish leadership" turns them down. His statement followed a week in which 40 of 41 local ballot…
1 sourcefinance.yahoo.com -
Nilay Patel criticizes AI industry’s products and costs in election aftermath
Nilay Patel, writing for The Vergecast, argues that AI companies are failing to justify their high costs and poor product quality. He claims current offerings—including AI agents and deepfake content—do not deliver…
1 sourceSimon Willison -
Researchers test Cross-Agent formalization for clinical trial criteria
A new study on arXiv explores how two AI models—one acting as a Specifier and the other as an Implementer—can independently translate clinical trial criteria into formal logic using the Dafny programming language. The…
1 source primary sourcearXiv cs.AI -
OpenAI leaks GPT-6.2 Bell model with 10 trillion parameters, 87% benchcad 2.0 score
The piece reports that OpenAI's next model, codenamed "Bell", was leaked in late October and early November 2026. The leak includes a 10 trillion‑parameter architecture and a reported 87% score on BenchCAD 2.0, a…
1 sourcenote.com -
Accountants face rising AI-briefed clients with flawed tax advice
Service firms like accounting are struggling as clients arrive with AI-generated advice—often incorrect—expecting professionals to implement it. OpenAI data shows U.S. tax-related ChatGPT queries quadrupled from Q1…
1 sourceUnite.AI -
Point One Navigation’s FUSE 2026 event announces robotics platform built on GPS accuracy
Point One Navigation, a robotics startup co-founded by Aaron Nathan, will showcase its FUSE 2026 physical AI event on November 19 in San Francisco. The company, led by Nathan—a veteran in robotics, sensor fusion, and…
1 sourceThe Robot Report -
Researchers propose TRAP policy for AI agent authorization
A team of researchers introduced the Trust Responsive Authorization Policy (TRAP) in a new arXiv paper. TRAP dynamically adjusts an AI agent’s operational authority based on a Trust Readiness Score (ARS), which blends…
1 source primary sourcearXiv cs.AI -
XIIlab approves KRW 10B convertible bond for ai data center
XIIlab disclosed that its board approved a KRW 10 billion privately placed convertible bond for its AI data‑center business. The bond carries a 0 % coupon, matures on November 16 2031, and can be converted at KRW 8,106…
1 sourceUnite.AI -
Square Peg backs Future Secure AI to build enterprise AI co-workers
Venture firm Square Peg announced on November 6, 2026, that it is backing Future Secure AI, a company building enterprise AI co-workers designed to complete complex business processes inside customers' existing…
1 sourceUnite.AI -
TSMC reports 41.2% revenue growth but sees margin pressure from 2nm ramp
Taiwan Semiconductor Manufacturing Company (TSMC) posted third-quarter earnings on October 15, with revenue of $46.72 billion—up 41.2% year-over-year—and earnings per ADR of $4.60, exceeding the $4.46 consensus…
1 source247wallst.com -
AI accidentally deletes professor's notes
| A University of Chicago law professor faced a concerning incident involving Claude AI, an AI model tasked with updating page numbers in lecture notes. | While the files were later recovered via Dropbox, the incident…
1 sourceTom's Hardware -
FlexFloor improves reinforcement learning post‑training efficiency for LLMs
A new system called FlexFloor tackles a major bottleneck in reinforcement‑learning post‑training for large language models. The approach releases compute workers earlier during synchronous response generation by using…
1 source primary sourcearXiv cs.AI -
Open ASR Leaderboard adds Egyptian and Nigerian English accents
The Open ASR Leaderboard has expanded its private data sets to include two new accents: Egyptian and Nigerian English. This addition follows a previous collaboration with Voice Arena for Indian-accented English and…
1 source primary sourceHugging Face -
Tau Robotics offers humanoid robot cleaning service in San Francisco
Tau Robotics launched a pilot cleaning service in San Francisco using four‑and‑a‑half‑foot humanoid robots. The company charges $30 for an hour‑long cleaning with a single robot, $60 per weekly cleaning with two…
1 sourceMIT Technology Review AI -
AI agents design and run robot labs in unreviewed preprint
A team of researchers has developed a system of artificial intelligence agents that can design, simulate, and operate robotic laboratories. Described in an unreviewed preprint posted to arXiv in October, the system…
1 source primary sourceNature Machine Learning -
Microsoft Foundry Local runs Qwen3.8-27B locally with GitHub Copilot on Nvidia GPUs
Microsoft released Foundry Local CLI 0.13.0, which runs the Qwen3.8-27B model on Nvidia GPUs and connects it to GitHub Copilot via Bring Your Own Key. The 4-bit quantized model needs about 18 GB VRAM, so it fits on…
1 sourcedevblogs.microsoft.com -
Zapier’s Sidekick shifts AI approvals from tools to protected resources
Zapier’s Sidekick AI workspace has refined how it handles user approvals for automated actions. Early versions asked for permission on a per-tool basis, but this led to confusion and inconsistencies. The team now ties…
1 sourceZapier Blog -
Researchers introduce BCSS to scale symbolic models of reasoning
A new paper on arXiv introduces Behavior-Conditioned Symbolic Synthesis (BCSS), a method designed to automate the creation of interpretable symbolic programs. The authors argue that traditional symbolic models are…
1 source primary sourcearXiv cs.AI -
Turkey reclaims 100 looted antiquities with AI and geochemistry from US
Turkey’s Culture Ministry returned 100 looted artifacts to Istanbul on Monday, including 44 terracotta panels from Duver (700–480 BCE) and 40 items from Horoztepe. The objects were recovered using lead isotope…
1 sourceartnews.com -
Teachers and parents: what are your experiences of AI-generated emails?
The survey targets teachers who have received AI-written messages from parents and parents who use AI tools like ChatGPT or Claude to draft emails to teachers. It seeks feedback on whether these emails improve or…
1 sourceThe Guardian AI -
Researchers develop gait foundation model to predict 3,210 health phenotypes
A new gait foundation model was trained on 3,328 adults who performed five motor tasks recorded with a single‑depth camera. The model learns embeddings that capture how all body joints move together over time, rather…
1 source primary sourceNature Machine Learning -
AWS shows how to govern AI agent access with Okta and Bedrock AgentCore Gateway
AWS published Part 2 of a tutorial on governing AI agent access using Okta Cross App Access (XAA) and Amazon Bedrock AgentCore. The post moves from a small topology—one agent and two Model Context Protocol (MCP)…
1 source primary sourceAWS Machine Learning Blog -
Vergecast discusses 2026 midterms and AI business challenges
The Vergecast hosts Nilay and David analyze the impact of the 2026 midterm elections on the technology sector. They note that voters favored Democrats and expressed opposition to AI, which may intensify regulatory…
1 sourceThe Verge AI -
Researchers propose KDASO for automated liability adjudication
A new paper on arXiv introduces KDASO, a framework designed to improve how Large Language Models handle liability adjudication tasks, such as legal charge determination. The authors argue that current LLM approaches…
1 source primary sourcearXiv cs.AI -
Microsoft renames GitHub Copilot harness to Microsoft Copilot harness, expands it across products
On October 28, 2026, Microsoft announced it is renaming the GitHub Copilot harness to the Microsoft Copilot harness and expanding it as a common foundation for its Copilot product family, including Copilot Studio,…
1 sourcenote.com
