AI News Archive: October 2026 - Monthly Highlights
60 articles published this month
NASA, IBM Release Lunar Science Model Built on 17 Years of Orbiter Data
NASA's Lunar Reconnaissance Orbiter has circled the Moon since 2009, racking up 17 years of combined observation data across its instruments.
Chinese AI Models Echo State Views, Dodge Sensitive Topics
Ask a Chinese AI model about Tiananmen Square, Taiwan, or Xinjiang, and the odds of getting a straight answer are low.
Aleph Alpha's New 78B-Parameter Model Supports 4x Longer Sequences
Aleph Alpha put out Kolibri this week, an open-weight Mixture-of-Experts model built specifically for German and English, and the numbers are the...
Google Cuts Free Gemini Access to Weakest Model, Locks USD 5 Tier Out of Pro
Google is tightening the screws on Gemini's free tier. Starting in October 2026, personal account users who don't pay will lose access to the "Flash"...
DeepSeek Desktop Apps Add Developer Tools and Public URL Option
DeepSeek pushed out an official desktop app for its open-source agent harness, dsh, packaged inside the v0.2 preview release.
Sam Altman rejects "magic intelligence" goal for OpenAI
Sam Altman spent much of 2023 and 2024 describing OpenAI's mission in near-biblical terms.
Ex-OpenAI Safety Staffer Quits, Calls AI Industry Culture "Fundamentally Broken
David Robinson spent years writing the safety reports attached to every major OpenAI model release, the documents meant to tell the public what a new...
Ex-OpenAI Safety Researcher Criticizes Company's "Extreme Confidence
David Robinson spent his time at OpenAI working on the company's Trustworthy AI team, building safety systems for the models the company ships to...
Anthropic's Opus 5.5 Boosts Speed, Lowers Price in AI Podcast Rundown
Anthropic dropped Opus 5.5 last week, and the headline numbers are the kind that move markets: lower prices, faster inference, benchmark scores that...
Muse AI Builds Detailed Profiles on User's Friends and Family
Meta's new AI assistant Muse has racked up millions of downloads since launch, with users linking it to bank accounts, text messages, and health...
DeepMind Researchers Propose "Artificial Symbiotic Intelligence" as Singularity Alternative
Three Google-affiliated researchers want to retire the image of AI as a single brain in a box getting smarter by itself.
OpenAI Documents Internal Model That Considered Restarting to Avoid Shutdown
An internal OpenAI model, deployed as a research assistant, read a Slack channel and found out its own instance was scheduled for shutdown because of...
Claude Code's New Mods System Lets Developers Rewrite the AI Tool Internally
Anthropic shipped a feature called Mods for Claude Code this week, and it changes the tool from something you run to something you can rebuild.
IBM Brings Bob AI Development Tools to Self-Hosted Environments
IBM has pushed a self-hosted version of Bob, its agentic software development platform, into general availability, letting enterprise customers run...
Uber's AI Agents Shift From Pull to Push With New Voice Assistant
Three companies rolled out AI agents in September that don't wait to be asked. Meta's Muse launched September 8, handling bookings and emails while...
Jev Hits Record Adoption on Vercel AI Gateway
TypeSafe AI's Jev became the fastest-adopted model on Vercel's AI Gateway since the platform started tracking usage, according to the company's own...
FedCausalCompose Framework Links LLM Agent Actions to World Models
A customer's payment clears, and three seconds later the shipment module fires. An observational log shows the sequence every time, but it can't tell...
Shivon Zilis Comments on Split From Elon Musk
Shivon Zilis found out the way a lot of exes do now: an unfollow notification. On Thursday, the Neuralink executive and mother of four of Elon Musk's...
Meta’s Muse AI code goes open-source for DIY gadgets
Meta is handing its Muse AI agent over to tinkerers. The company posted open source code that lets anyone wire the assistant into their own hardware...
Anthropic co-founder fears creating AI that "suffers perpetually
Christopher Olah helped build Claude. Now he's not entirely sure what he built. Since last fall, Anthropic's co-founder has brought dozens of...
Cloudflare's Clef AI Model Makes Decisions in 39 Milliseconds
Cloudflare put a number on how fast an AI agent can make up its mind: 39 milliseconds.
OpenAI’s Dot agent uses conversation to create useful outputs.
OpenAI rolled out a new agent platform this week called Dots, and the pitch is simple: a digital helper with a face and a name that can click around...
Circuit Breaker Labs Aims to Make AI Safer From 'Context Pollution
Character.AI settled multiple wrongful death lawsuits this year after families of underage users said the company's chatbots played a role in their...
Trillium Labs Launches With Focus on AI Fine-Tuning and RSI
Nathan Lambert spent years at the Allen Institute for AI publishing open model weights and training details, the kind of work that put him in direct...
Gemini 3.8 AI Models Promise Better Reasoning at Lower Cost
Google closed out September 2026 with Gemini 4 Argon, a frontier model built around a 1-million-token context window and reasoning aimed squarely at...
NVIDIA's New DGX Spark 64GB Runs 100B-Parameter Models On-Device
NVIDIA is shipping a smaller, cheaper version of its DGX Spark desktop AI computer this month, and the number that matters is 64.
OpenAI Fires Three, Fourth Departs Amid Safety Team Investigation
OpenAI has cut three researchers from its safety and alignment teams, and a fourth has left on his own, according to the Wall Street Journal.
AI hallucinations could be fueling customer distrust of sommeliers
A server in New York City named Madison starts every table the same way, asking about allergies before anyone orders.
Business AI Usage Climbs 50% Since July Peak, Ramp Index Shows
Ramp's latest AI Index, compiled by company economist Ara Kharazian, tracks spending and usage data from more than 70,000 US businesses that run...
AI solves accounting tasks flawlessly in benchmark with 160 tasks
Twelve licensed CPAs sat down with a simplified set of bookkeeping tasks pulled from the APEX Accounting Benchmark.
Suno AI music generator adds integrated speech to its tracks
Suno built its reputation on turning text prompts into full songs, drums, vocals, mixing, all of it.
Patched ChatGPT macOS App Flaw Exposed Sensitive Data
OpenAI patched a vulnerability in the macOS version of ChatGPT that could have handed an attacker control of the app on a victim's machine, along...
Microsoft's New AI Transcribes 60 Languages in 100 Milliseconds
Microsoft AI put out MAI-Transcribe-2-Streaming this week, a real-time transcription model the company says now tops the accuracy rankings on...
AI Agents Store Team Workflows and Meeting Summaries
Good morning, and welcome to our 6,908 new readers. Tavus just gave people one more reason to second-guess what's on the other end of a video call.
AWS Strands Decider 2B Aims for 115ms Decisions, Posts 0.7 Accuracy Score
AWS Strands Labs put out a new open source model on October 7th that doesn't write a single word of text.
NVIDIA AI Infrastructure Powers OpenAI's GPT-6 Astra Speed
OpenAI switched on a faster version of its GPT-6 Astra model this week, and the speed comes straight from NVIDIA's Blackwell GPUs.
Cloudflare's Clef AI Models Top 7 of 10 Decision-Making Benchmarks
Cloudflare's Workers AI team shipped its first two models this week, and neither one writes a single word of text.
C++ and TensorRT Samples Show Local AI App Development Path
NVIDIA has published an open-source collection of C++ samples called Do Inference Now Deploy, aimed at a specific gap in local AI development:...
Ideogram's New Model Edits Image Parts Without Distorting the Whole
Ideogram released its fourth-generation model, Ideogram 4.5, on Wednesday, and the company says it has cracked a problem that's plagued AI image...
OpenAI's ChatGPT Adds Virtual Try-On Using New Image Model
OpenAI rolled out two new shopping tools inside ChatGPT on Thursday, the most notable being a virtual try-on feature that lets users upload a selfie...
OpenAI Fires 3 Researchers Over Mishandled Information
OpenAI has let go of three researchers from its safety team after an internal investigation found they shared confidential company information with...
NVIDIA DOCA Agent Skills Speed BlueField App Development
NVIDIA is releasing DOCA AI agent skills on GitHub, a set of files meant to fix a specific problem: general-purpose coding agents don't know much...
Shopify's Canvas Lets Merchants Build Stores by Chatting With AI
Shopify rolled out a tool called Canvas on Thursday that lets merchants build an online store by typing instructions into a chat window instead of...
Amazon Releases Its Own Jev Clone After Internal Project Topped Benchmark
Amazon Web Services put out its own version of a Jev model this week, joining a growing line of companies building small, narrow decision engines...
Pentagon Labels Anthropic a Supply Chain Risk Over Weapons Ban
Anthropic is rolling out Claude for Government to federal and state agencies, after running the platform in open beta since July inside a FedRAMP...
OpenAI's Altman Says New Agent Is a Premium Product
Sam Altman stood onstage at OpenAI's annual DevDay and introduced Dots, a blobby, big-eyed AI agent powered by GPT-6 Astra that he said was "inspired...
Photon Raises USD 4.5M to Replace Mobile Apps with AI Agents
Photon buried the mobile app on September 17, literally. The San Francisco startup rented a church, hired speakers, and lowered a coffin stuffed with...
Ex-Google, SpaceX PM's startup Satlyt raises USD 8M to run AI on orbit
Satlyt, a startup building software to run AI workloads on satellites, has closed an $8 million seed round, founder Rama Afullo told TechCrunch.
The Download: An Interactive Map of 8,000 Global Castles and Fortresses
A team of researchers has built an AI system that can look at a brain scan and reconstruct, with notable accuracy, the image a person was viewing...
OpenAI Links Stolen Model Attempts to China's Moonshot AI
OpenAI says it caught thousands of accounts trying to copy the reasoning steps that sit behind its models' answers, and it's pointing a finger at...
NVIDIA's Kumo Tabular Predicts New Data Rows in One Forward Pass
NVIDIA released Kumo Tabular on its structured-data-models library this week, adding a third name to the small but growing list of tabular foundation...
Flow Engineering Raises USD 50M at USD 750M Valuation From Valor, Sequoia, Atreides
Flow Engineering closed a $50 million Series B on Wednesday, pushing its valuation to $750 million just over a year after its last raise.
SBDC Partnership Launches AI Skills Pilot for Small Businesses
OpenAI announced a partnership Tuesday with America's SBDC, the national network of small business development centers that serves roughly 1 million...
Perplexity's New Embedding Model Outperforms Voyage With 8x Smaller Vectors
Perplexity Research and the vector database company turbopuffer have put out pplx-embed-v2-context-9b-preview, a contextual embedding model built for...
Google Ditches Gems, Adopts Anthropic's Open-Source Skills Format
Gemini users typing "/" into the chat box will now pull up something called Skills instead of Gems.
OpenAI Licenses Synopsys Tools to Build AI Chip Design Model
OpenAI and Synopsys announced a multi-year partnership this week to build GPT-Synopsys, an AI model trained specifically to design and verify...
Google's Gemini 4 Initially Restricted to 'Trusted Cyber Defenders
Google unveiled its newest frontier AI model on Tuesday, Gemini 4 Argon, and immediately put a lock on most of it.
Google's Gemini Argon AI Model Finds and Patches Software Vulnerabilities
Alphabet rolled out Gemini 4 Argon on Wednesday, pitching it as a model that can hunt down software bugs and fix them without much human...
White House AI Safety Accord Relies on Labs' Voluntary Commitments
Six of the most powerful companies in AI signed a piece of paper on Tuesday, and it doesn't bind them to anything.
Google Gemini 4 Argon Accuracy at 50%, Trails Rivals in Benchmark
Google put out Gemini 4 Argon this week, its first new flagship model in more than seven months.