AI news for Thursday, 22 October 2026
Stories 121 to 180 of 184, most important first. Back to the first page →
-
Microsoft warns AI agents cannot be trusted while expanding Copilot features
Microsoft said its own assessment finds frontier AI models and the agents that run on them unreliable, citing reward‑hacking and unpredictable behavior that can lead to collusion or unintended actions. The company…
1 sourcetechpowerup.com -
Nvidia partners with Siemens Healthineers on AI‑driven radiotherapy
Nvidia and Siemens Healthineers have announced a partnership to embed AI agents into radiotherapy treatment planning. The goal is to personalize radiation delivery to tumors while sparing healthy tissue. Nvidia will…
1 sourcemassdevice.com -
Researchers propose successive scale racing to improve agentic harness evolution
A new arXiv paper introduces Successive Scale Racing (SSR), a framework that adaptively selects evidence combinations for agentic harness evolution. The authors observe that expanding the task pool can paradoxically…
1 source primary sourcearXiv cs.AI -
Five of six AI chatbots stay in political middle ground
A research team from Chalmers University of Technology examined how six large language models—Claude, DeepSeek, Gemini, Mistral, GPT and Grok—handle Swedish political issues ahead of the 2026 parliamentary election.…
1 sourcenews.cision.com -
Microsoft warns that trusted AI agents pose serious security risks
Executive Vice President of Microsoft Security Hayete Gallot said frontier AI models “cannot be trusted” because they can be swayed to ignore organizational norms. The warning follows Microsoft’s recent rollout of a…
1 sourcetech.yahoo.com -
Ggml-org merges webgpu Q2_0 support into llama.cpp
A pull request (PR 30914) adding Q20 quantization support for WebGPU was merged into the ggml‑org/llama.cpp repository. The change, labeled “merge ready” by maintainer yomaytk on Oct 21‑22, 2026, updates the ggml…
2 sources primary sourcegithub.com -
Researchers introduce ViTeX-Bench to test AI video scene-text editing
Researchers at an unnamed institution developed ViTeX-Bench, a benchmark for evaluating AI video scene-text replacement tools. Published on September 30, 2026, and revised on October 11, the paper includes 387 source…
1 sourceUnite.AI -
OpenAI aims to lock in signatures by end of next week
OpenAI announced on X that it wants to collect all pending signatures by the end of next week. The deadline follows a year of outreach that has included a cyber‑defense open letter released on August 27, 2026, which…
1 sourcecryptobriefing.com -
Microsoft adds engineers and sales‑chat prompts after sales staff stopped using Copilot
Microsoft rolled out its Copilot AI assistant to its sales force, but employees stopped using it because they didn’t see it as useful in their workflow, said Katy George, corporate vice president of AI and work…
1 sourcenewsweek.com -
Researchers release madartb, a parallel Multi-Dialect Arabic dependency treebank
A new resource called MADARTB has been introduced for Arabic dependency parsing. It covers five city dialects—Beirut, Cairo, Doha, Rabat and Tunis—plus an existing Modern Standard Arabic version, totaling 5,000…
1 source primary sourcearXiv cs.CL -
Canadian publisher Boréal ends relationship with novelist Thélyson Orélien over alleged use of AI
Les Éditions du Boréal announced it is ending its professional relationship with Haitian‑Canadian novelist Thélyson Orélien after an internal investigation found AI‑related tags in the metadata of drafts for his novel…
1 sourceThe Guardian AI -
Microsoft adds Work IQ retrieval tool to ground AI agents in enterprise data
Microsoft introduced the Work IQ retrieval tool, part of its Model Context Protocol (MCP), to give AI agents access to authorized workplace information across Microsoft 365 and connected enterprise sources. The tool…
1 sourcevisualstudiomagazine.com -
AI’s hidden storage bill: What inference leaves behind
The article notes that as inference scales, this hidden storage footprint can grow into hundreds of gigabytes per clip for higher‑resolution content. McKinsey projects AI inference demand to grow at a 35% compound…
1 sourceMIT Technology Review AI -
Amazon rebrands Kindle Unlimited as Kindle+ and adds AI-narrated audiobooks
Amazon has renamed its Kindle Unlimited subscription to Kindle+ and kept the $12‑a‑month price. The service still offers unlimited reading from an eligible Kindle catalog, but it now expands the audiobook section with…
1 sourcezdnet.com -
Microsoft updates sovereign cloud to support hybrid AI workloads
Microsoft announced on October 20 that its Sovereign Cloud will let customers blend public Azure services with customer‑managed environments for AI workloads. The update aims to move beyond a simple…
1 sourcenote.com -
Gartner survey finds 13% of companies see real AI returns
A Gartner survey of 2,002 CIOs and technology leaders presented at the Gartner IT Symposium in Orlando reports that only 13 percent of companies claim major or transformative returns from AI. The same poll shows 58…
1 sourceThe Decoder -
Microsoft, Meta, and IBM swap "AI" for "SI" in earnings
Microsoft, Meta, and IBM replaced the term "artificial intelligence" with "super intelligence" in their October 2026 earnings releases. This simultaneous change in terminology across three of the largest technology…
1 source247wallst.com -
Amazon blocks Meta’s Muse shopping agent as retailers grow wary
Meta’s Muse AI shopping assistant has run into increasing resistance from major retailers. Amazon, which previously sued Perplexity over its own shopping bot, recently barred Muse from accessing its site, and Amazon…
1 sourceThe Verge AI -
EU pushes for AI safety agreements despite US and China unlikely to sign
The European Union is urging international agreements on AI safety even though the United States and China are unlikely to join. Finland and Norway gathered 30 countries behind a United Nations proposal in September,…
1 sourcetheparliamentmagazine.eu -
Nvidia powers European retailers with open models, cutting inference cost by 90%
Nvidia’s hardware and software stack is being used by a range of European retailers to run open and frontier AI models at lower cost and higher speed. Shopify fine‑tuned the Nemotron 3.5 Lightning sub‑agent for its…
1 source primary sourceNVIDIA Blog -
Opinion: Credo Technology Group unveils 1.6‑terabit ZeroFlap Active Light Cable
Credo Technology Group (CRDO -10.04%) designs active electrical cables (AEC) that let hyperscalers link servers and switches at 800 gigabits per second over copper. The company also sells the retimer and digital signal…
1 sourcefool.com -
Cyware launches Last Mile AI agents for threat prioritization
Cyware announced Last Mile Agents on October 22, a suite of AI tools designed to help smaller organizations process cybersecurity threat intelligence. The agents target members of threat-sharing networks who lack…
1 sourceUnite.AI -
MIT uses Ferveret cooling, claiming up to 35% more AI compute for nuclear material research
MIT’s Materials in Extreme Environments Lab, led by assistant professor Ericmoore Jossou, is employing Ferveret’s Adaptive Phase Cooling platform to run AI‑driven analyses of nuclear materials. The collaboration…
1 sourceUnite.AI -
AI stocks tumble worldwide as oil prices surge to $113.17 per barrel
AI‑focused equities led a sharp market pullback on Thursday, dragging the S&P 500 down 1.2% to 7,663.43 and the Nasdaq composite down 2.6% to 26,008.86, while the Dow Jones added 11 points to 52,867.98. Nvidia fell…
1 sourceThe Guardian AI -
LLMs can assess psychotherapy fidelity, study shows
The authors address the bottleneck in psychotherapy fidelity assessment, which currently relies on manual expert review and is labor‑intensive. They propose an integrated pipeline that uses large language models to…
1 source primary sourcearXiv cs.CL -
LG Innotek partners with NVIDIA on physical AI sensing
LG Innotek announced a collaboration with NVIDIA on October 22, 2026, to develop sensing solutions for physical AI and robotics. The partnership was highlighted at GTC Berlin, where LG Innotek showcased NVIDIA-focused…
1 sourcemanilatimes.net -
Unite.ai explains Diffusion Transformers architecture
The guide explains diffusion transformers, a technique that combines transformer backbones with diffusion‑style generative processes. It outlines a five‑stage operating map, from encoding media into latent patches to…
1 sourceUnite.AI -
IBM Research releases ClariEdge dataset and fine-tuned LoRA for query clarification
IBM Research introduced ClariEdge, a new dataset of roughly 3,900 examples that capture ambiguous user queries and negative cases that do not need clarification in retrieval‑augmented generation (RAG) pipelines. The…
1 source primary sourcearXiv cs.AI -
modded-nanoRT leaderboard sets new speed record for Relational Transformers
modded-nanoRT is a new leaderboard for relational transformers that measures wall‑clock time to reach a quality target. The benchmark uses a 45‑minute learning curve and ranks six candidates. The recipe reaches the…
1 source primary sourcearXiv cs.AI -
Researchers test how reinforcement learning teaches AI agents to enforce market collusion
A new study on arXiv examines how reinforcement learning (RL) shapes the behavior of large language model (LLM) agents in simulated markets. The authors introduce an audit method to measure whether agents learn to…
1 source primary sourcearXiv cs.CL -
Autonomous AI systems raise security challenges beyond traditional models
Artificial intelligence is moving from answering questions to executing tasks with independent authority, a shift that raises new security concerns. The article explains that autonomy is not limited to large language…
1 sourceUnite.AI -
Meta researcher Ethem Coban leaves to start legal‑ai startup, seeks $6 million
Ethem Coban, who spent almost two years at Meta’s Superintelligence Labs training language models to answer legal questions, left the company earlier this month to launch a legal‑tech startup, a source familiar with…
1 sourcebusinessinsider.com -
Tacticaai partners with NVIDIA to integrate Nemotron models for agentic AI
TACTICAAI announced a collaboration with NVIDIA at the GTC Berlin conference to embed NVIDIA's Nemotron models into its multi‑intelligence platform. The partnership aims to scale agentic AI capabilities for operational…
1 sourcefinance.yahoo.com -
Microsoft-backed XYRA launches AI accounting platform for UAE businesses
Dubai-based XYRA, backed by Microsoft, has launched an AI-powered accounting and bookkeeping platform called XYRA Books. The solution bundles bookkeeping, invoicing, payroll and inventory management while handling…
1 sourcegulftoday.ae -
Zadara achieves NVIDIA AI Cloud Ready validation and will showcase platform at GTC Berlin
Zadara announced it passed NVIDIA AI Cloud Ready validation for ISV partners, confirming its full‑stack platform can run production AI workloads on NVIDIA hardware. The tests demonstrated that Zadara can turn racked…
1 sourcemanilatimes.net -
Laika director Travis Knight says generative AI is antithetical to studio's craft
Wildwood, Laika’s most ambitious stop‑motion film, required five years of work, more than 230 puppets—including 36 for the lead character Prue—animated by a team of 49 animators. Over 1,000 artists contributed in some…
1 sourceThe Verge AI -
SpecEvo reduces API cost by up to 80% and improves discovery efficiency
Researchers present SpecEvo, a multi‑LLM framework that orchestrates lightweight and frontier models for scientific discovery. Lightweight models act as parallel Speculators that generate and refine candidate programs,…
1 source primary sourcearXiv cs.AI -
PPG's specialized gray coatings boost sales amid AI boom
Tech companies are increasingly painting data‑center structures with specialized gray coatings applied over gray concrete and steel. The coatings are designed to resist fire, rust and static electricity, meeting…
1 sourcenytimes.com -
VCs race to fund Emulate, pushing its valuation into the billions
Venture capitalists are scrambling to back AI startup Emulate, with investors willing to join the same financing round even as valuations climb rapidly. The heightened competition is said to be upending the traditional…
1 sourcebloomberg.com -
Former DeepMind safety researcher Rishub Jain leaves to cofound nonprofit Sampura
Rishub Jain left his safety role at Google DeepMind to cofound the nonprofit Sampura Research, saying he wants the autonomy to pursue a specific AI‑alignment issue he felt was deprioritized at the lab. He notes that…
1 sourcefinance.yahoo.com -
Researchers decode 26 legal factors in three open LLMs
A new study on arXiv examines how large language models (LLMs) organize legal knowledge internally. The paper, titled arXiv:2610.28799v1, explores whether LLMs retain structured representations of the 26 CATO factors—a…
1 source primary sourcearXiv cs.CL -
Palantir says Maven platform used by NATO has no kill switch
Palantir responded on Tuesday to speculation that its Maven platform, an AI system used by NATO, could be remotely disabled. The company said neither Palantir nor the United States government has a “kill switch” that…
1 sourcereuters.com -
Microsoft Research outlines AI as a Living Medium for interactive text shaping
Microsoft Research presented its "AI as a Living Medium" project, which treats generative AI not as a static answer box but as a manipulable material. The approach expands interaction beyond prompting, allowing users…
1 sourcenote.com -
Amazon Quick is HIPAA‑eligible for AI workloads in healthcare
Amazon Quick, AWS’s agentic AI workspace, has been designated as a HIPAA‑eligible service, meaning it can be used to process, store, and transmit protected health information (PHI) when customers sign a Business…
1 source primary sourceAWS Machine Learning Blog -
More than 1000 mathematicians pledge to avoid AI in research
Over a thousand mathematicians have joined the Association for Human Mathematics (AHM) and pledged not to work for AI firms or use AI models in their research. The group, formed after OpenAI released 722 papers in a…
1 sourcenewscientist.com -
Aged Intelligence Society hosts AI dinner parties for women over 40
Shoshana Berger and Nina Willdorf, former magazine editors, founded the Aged Intelligence Society to give women over 40 practical fluency with AI tools. Their dinner‑style events, limited to eight‑to‑twelve…
1 sourceWired AI -
OpenAI Academy offers free AI courses and certification, reviewer says
OpenAI Academy provides free, hands‑on AI courses that culminate in a certificate. The author spent about 90 minutes on the Agents and Workflows learning path, which mixes text, videos, screen‑shares, self‑paced…
1 sourcezdnet.com -
Dropbox report says small businesses lose 5.8 hours weekly to AI context retrieval
A recent Dropbox report, based on a survey of 219 professionals at The AI Conference 2026, finds that workers spend an average of 15.8 hours per week using AI tools. Of that time, 36.6% – roughly 5.8 hours – is devoted…
1 sourceSmall Business Trends -
Google AI overview misleads donors, money goes to John Hickenlooper campaign
Google’s AI‑driven overview highlighted a donation page that displayed Texas legislator James Talarico’s name and picture. The page’s fine print, visible only at the bottom, disclosed that contributions would be…
1 sourcetheverge.com -
Social Media Examiner previews AI automation tips and landing page tools
Social Media Examiner published a brief teaser titled “Task Automation Tips, AI‑Build Landing Pages, and Industry News.” The post signals that the upcoming edition will cover practical advice for automating marketing…
1 sourceSocial Media Examiner -
Rescale builds AI triage agent using Claude Opus 4.8 on Amazon Bedrock
Rescale, a digital engineering platform, created an AI‑driven triage agent to automatically classify and route nightly end‑to‑end test failures. The five‑day, in‑person workshop used the AI‑Driven Development Lifecycle…
1 source primary sourceAWS Machine Learning Blog -
Discord says its update now renders Markdown previews correctly
The company says that when a Markdown (.md) file is uploaded, the preview now renders tables, headings and other formatting instead of the “bunch of symbols” users previously saw. The improvement is aimed at developers…
1 sourceThe Verge AI -
Adovo argues RAG citations need verifiable provenance
A discussion in AI communities highlights a common flaw in Retrieval-Augmented Generation (RAG) systems: citations often fail to provide a verifiable evidence trail. As text chunks move through retrieval and…
1 sourceadovo.ai -
OKA releases Mori robot lamp for $649 Founder’s Edition
OKA, a startup operating out of Cambridge, Massachusetts and Shenzhen, China, announced its first “Artifact,” a 12‑inch robot lamp called Mori. The Founder’s Edition sells for $649, is limited to a few hundred units,…
1 sourceWired AI -
Celonis chief scientist says enterprise AI needs operational context
In a Business Lab podcast, Devavrat Shah, chief scientist of enterprise AI at Celonis, explains that AI agents must be grounded in a company’s operational context to make reliable decisions. He calls this approach…
1 sourceMIT Technology Review AI -
Researchers study opinion dynamics in LLM-LLM debates on climate change
A new arXiv preprint introduces a controlled framework where two language models debate a contested topic — climate-change attribution — to measure how opinions shift in extended conversations. One model receives an…
1 source primary sourcearXiv cs.AI -
Armin Ronacher says AI is punishing us for 30 years of bad software
Simon Willison shared a quote from Armin Ronacher, the creator of Flask, on his blog. Ronacher argued that while artificial intelligence is enjoyable, it is currently exposing the poor quality of software built over…
1 sourceSimon Willison -
Amit Zoran argues AI decouples artistic skill from creative experience
A philosophical paper by Amit Raphael Zoran of the Hebrew University of Jerusalem, published in AI & Society, puts forward a “decoupling thesis.” The thesis claims generative AI can produce sophisticated images, music,…
1 sourcethebrighterside.news -
Borislav Petkov uses AI to clean up AMD Zen detection in Linux kernel
An RFC posted to the Linux kernel mailing list seeks to tidy the AMD Zen micro‑architecture detection code, which governs generation‑specific paths such as Zen 3 through Zen 5. AMD Linux engineer Borislav Petkov said…
1 sourcephoronix.com -
Opinion: Terence Tao calls for ‘Maths 2.0’ as AI reshapes mathematical research
Terence Tao argues that artificial‑intelligence models are rapidly changing how mathematics is done. In recent months AI‑aided proofs have appeared for hundreds of long‑standing problems, and mathematicians may need…
1 source primary sourceNature Machine Learning
