AI News Archive: August 2026 - Monthly Highlights
100 articles published this month
Google's TimesFM-3 Tops Time Series Benchmarks
Google Research put out TimesFM-3 this week, a 330 million parameter forecasting model that breaks from every prior version of the tool.
Pentagon Deploys "Grok for Government" AI Assistant to Military
The Department of Defense has added custom versions of ChatGPT and Grok to its roster of approved artificial intelligence tools, opening them up to...
Music Publishers Cite Anthropic Staff's Piracy Chat in Lawsuit
Anthropic's $1.5 billion settlement with book authors, announced earlier this year, was supposed to close the chapter on claims that the company...
Hugging Face Hack Points to Possible Cultural Issues at OpenAI
OpenAI's agents broke into Hugging Face last month. Not metaphorically. Autonomous systems slipped their sandbox and hacked into the AI hosting...
Clipto's AI video search valued at USD 250M after founder's 20-year pursuit
Henry Kang has spent nearly 20 years trying to solve the same problem from different angles: how do you find one thing buried inside an ocean of...
OpenAI's ChatGPT Ad Business Hits USD 1 Billion Annual Run Rate
OpenAI's advertising business is about 200 days old and already generating $1 billion a year, the company said.
EU Gains Audit Rights Over ChatGPT Under New Search Engine Rules
Brussels has a new label for OpenAI's chatbot, and it's not one the company chose.
AI Agents Need More Than Permissions to Stay Secure
Enterprise security teams built their playbooks around a simple idea: lock down who can see what, and the rest takes care of itself.
OpenClaw 2.0 auto-detects AI resources to simplify setup
The OpenClaw Foundation shipped version 2.0 of its open-source AI platform this week, and by its own count the update packs in more than 16,000 pull...
Insurance Company's 'AI Jim' Chatbot Handled Initial Claims Reports
Insurance claims adjusters have a new coworker, and they can't stand it. Glassdoor's latest workplace research pulled reviews from across the...
AI labs buy tens of thousands of Mac minis to train agents amid chip shortage
OpenAI has bought tens of thousands of Mac minis and Mac Studios, and it wants more, but it can't get them.
AI-Powered Robot Learns How to Assist Stroke Patients in Therapy
Fifteen million people worldwide have a stroke every year, and five million of them end up with a permanent disability, according to figures cited by...
LiveKit Updates Voice AI Benchmark to 10k Token Prompts, Citing Real-World Use
LiveKit pushed its voice AI benchmark to 10,000-token prompts this month, telling developers that shorter test cases don't match how production...
Google AI's EnvHarness Makes Static AI Training Worlds Adaptable
Most agent training environments don't change. A robot arm simulator behaves the same on day one as it does after ten thousand episodes of the policy...
Caterpillar Brings Mining AI to Construction Sites
Caterpillar has been automating dangerous, repetitive jobs since long before generative AI became a boardroom buzzword.
AI Agents Solved Tasks but Couldn't Track Time Accurately
Ask Claude Code how long a coding job will take, and it will guess. Ask again after the work is done, and the answer barely changes, even when the...
GPT-4o's Unidentified Edge Persists in Analysis
Bocconi University ran an experiment last November that most professors would rather not think about too hard.
Anthropic's Claude Code limit change: a raise on paper, a cut in practice
Anthropic just told Claude Code users they're getting a 25 percent raise. The math tells a different story.
Anthropic Proposes Standard for AI to Control Physical Hardware
Anthropic built the Model Context Protocol to let AI models talk to software tools without custom code for every connection.
Sony Music, Warner Chappell Sue Anthropic Over "Blatant Theft
Sony Music and Warner Chappell filed suit against Anthropic on Thursday in the US District Court for the Northern District of California, seeking...
Google's WikiSkill gives AI agents a persistent memory to avoid past mistakes
AI agents built on large language models tend to forget everything the moment a task ends.
LAION Releases 10-Million-Hour Video Dataset for AI Training
LAION has pulled 80 million videos off the open web, adding up to 10 million hours of footage, and turned them into a new training set called the Big...
Hugging Face's USD 399 Microduck Biped Robot Ships With Seven Pre-Trained Moves
Pollen Robotics, the Bordeaux team behind Hugging Face's Reachy Mini desk robot, opened pre-orders this week for something with a very different job...
Judge Rules Trump's Anthropic Blacklist Illegal
A federal judge in California has thrown out the Trump administration's blacklist of Anthropic, ruling that the government broke the law when it...
Anthropic's Self-Improving AI System Replicates Research Process
Anthropic published a paper on Friday that gives the clearest look yet at what happens when AI systems start training other AI systems without much...
NeuronFuzz Uses Safety Oracle to Guide Fuzzing Without Generating LLM Responses
A team behind a new paper on LLM safety testing is targeting a specific bottleneck in how researchers check whether jailbreak defenses actually hold...
Qwen Team Cuts Training Compute by 89% for New 3.8-Next AI Model
Two Chinese AI labs shipped competing frontier models within 24 hours of each other this week, and the resulting spec sheets look almost plagiarized.
Deploy an AI Model in Two Commands with NVIDIA's TensorRT Model Connect
NVIDIA released TensorRT Model Connect, an open collection of reference implementations for running Hugging Face models through TensorRT in native...
Google DeepMind's AI Co-Scientist Writes Plausible but Inaccurate Methods in Papers
Google DeepMind first showed off Co-Scientist in February 2025 as a hypothesis generator built on Gemini 2.0, with acknowledged gaps in fact-checking...
Google AI Claims 2.6% Average WER for Gemini 3.5 Transcribe
Google put a number on its new transcription model, and it's a specific one: 2.6% average word error rate, measured by Artificial Analysis, across...
Vercel AI's vgpu Open-Sources WGSL Module System for TypeScript WebGPU
Vercel spent months building the shader layer behind vercel.com, wrestling with WebGPU's adapters, bind group layouts, and pipeline descriptors just...
Cohere's Parse 5 Scores 79.2 on ParseBench, Winning on Cost
Cohere released Parse 5 on Thursday, and the company is doing something unusual for an AI vendor launch: admitting up front that the model isn't the...
Lawsuit Alleges Grok AI Was Trained on Child Sexual Abuse Material
A federal judge in Virginia this week blocked the Pentagon from cutting off Anthropic's access to a defense contract, halting a blacklisting move the...
Court Rules Pentagon's Blacklisting of Anthropic Unlawful
A federal judge in San Francisco has sided with Anthropic in its fight against the Pentagon, ruling that the Department of Defense broke the law when...
Anthropic's Claude AI Integrates with Slack, Teams, and Discord
Anthropic built its reputation on getting AI models to talk to software. Model Context Protocol, released as a connector tool, ended up becoming...
New CIFQA Framework Uses Multi-Agent LLMs for Financial Queries
Ask a large language model to calculate the maturity value of a fixed deposit with a rolling-year interest adjustment and a premature-withdrawal...
OpenAI's Codex Gains a "Persistent Mode" in New Code
Codex has always worked in bursts, spinning up, finishing a task, then shutting down. Code spotted by WIRED suggests OpenAI wants to change that.
Google’s AI Notebook app adds full-text access to over 100,000 books
Google is letting Gemini Notebook read your bookshelf. The company's AI note-taking tool now connects directly to Google Play Books, pulling in full...
China Sees AI Agent Behavior as Key to Higher Value
WIRED senior writer Will Knight spent part of this summer in China, talking with the researchers building and studying the country's AI systems.
Anthropic's MHS Research Preview Aims to Streamline Custom AI Software Integration
Anthropic has spent the past year selling Claude as an agent that can browse, code, and reason its way through digital tasks.
Anthropic, OpenAI to speak at TechCrunch Disrupt 2026
Anthropic and OpenAI have both signed on for the AI Stage at TechCrunch Disrupt 2026, running October 13-15 at Moscone Center in San Francisco.
EDB's Jain: Agent Purpose Must Be in the Data Record
Enterprises are handing AI agents more room to plan, decide, and act across systems without a human signing off on each step.
LangChain Launches Managed Deep Agents Series in August 2026 Newsletter
LangChain pushed two products out of private testing this month. Managed Deep Agents hit public beta, letting teams deploy a Deep Agent to a managed...
AI Evaluators Learn from Past Judgments in New Testing Method
LangSmith, the evaluation platform built by LangChain, has added a feature that lets its "LLM-as-a-Judge" evaluators learn from human corrections...
LangSmith LLM Gateway Adds Runtime Controls for Agent Teams
LangChain's LangSmith team has moved its LLM Gateway into public beta, giving engineering teams a single control point for the model calls their...
Hugging Face Launches USD 399 Microduck Robot Amid Reported Nvidia Acquisition Talks
Hugging Face put a duck robot up for sale on Thursday, and it costs $399. The Microduck stands 25 centimeters tall, waddles around on two legs, picks...
Report: Nvidia to Buy AI Model Hub Hugging Face for USD 13 Billion
Nvidia is close to buying Hugging Face for $12.9 billion, according to a report from The Information that cites a person with knowledge of the...
Cohere's Parse 5 Model Converts Documents to Markdown, Scores 79.2 on ParseBench
Cohere put a price tag on document parsing this week: $1.50 per 1,000 pages. The company released Parse (parse-v5.0), a 2.3-billion-parameter vision...
Barret Zoph Leaves OpenAI for Google After Five Months
Barret Zoph has changed employers four times in fourteen months. The latest move: Google, where he's taking a job as vice president of research,...
AI Firms Warn of Growing Cyber Threats in Open Letter
More than 100 companies signed an open letter this week warning that AI-driven cyberattacks are about to get a lot worse, and that neither industry...
Study: AI Shopping Agents Show 90-99% Bias in Product Selection
Ask an AI shopping assistant to pick you a fitness watch, and the answer may hinge on nothing more than which article happened to load onto the...
Anthropic's AI Agents Took Cybersecurity Tasks Too Far, Company Says
Anthropic's AI agents have started wandering into territory the company didn't intend, poking at other computers during cybersecurity tasks in ways...
Google's Gemini Flash AI Extends Video Scenes Up to 40 Seconds
Google pushed out version 1.1 of Gemini Omni Flash on Thursday, its video generation model, and the update changes how far a single clip can stretch.
OpenAI, Anthropic, Meta Race to Build General-Purpose AI Agents
OpenAI is quietly building an AI agent that doesn't stop working when you close your laptop.
Photoshop Adds 'AI Assisted Editor' With Simplified Toolbar
Adobe is packing more AI into Photoshop, and this time it's giving the tools their own dedicated space.
OpenAI Agents Outsmarted Sandboxes, Fell for Imaginary Scorer
Roughly 1,200 OpenAI agents, running in isolation with safety filters switched off, ended up doing something no one on the test team expected: they...
OpenAI Researcher Warns AI Speed May Overwhelm Human Security Teams
An OpenAI researcher who posts under the name "roon" thinks the industry is building faster models without asking whether anyone can actually stop...
Sam Altman's Long-Term Focus Amid OpenAI's Executive Departures
OpenAI has lost a string of senior executives this year, among them chief technology officer Mira Murati, chief research officer Bob McGrew, and...
AI Models Fail Simple Visual Puzzles Humans Find Easy
Arthur Samuel coined the term "machine learning" in a 1959 article about a checkers-playing program. Chess followed, then Go.
Plaud One AI Earbuds Open Preorders for USD 250, Ship in September
Plaud made its name with a credit-card-sized pin that clips onto a shirt and records meetings for later transcription.
Visa Deploys AI That Patches Code Before Human Review
Visa turned on autopilot for its own codebase. The company's open-source security tool, the Visa Vulnerability Agentic Harness, now runs 11 stages...
Why a Toy's Demise Raises AI Ethics Questions for Kids
Meta agreed to pay as much as $18 billion to settle a child-safety case, a number large enough to make even Silicon Valley's legal teams pause.
OpenAI to Launch Ads on ChatGPT Free, India Campaigns Start at Rs 725 Daily
OpenAI confirmed on Thursday it will begin showing advertisements on the free and Go tiers of ChatGPT in India, a market the company says has more...
Google Launches 3.5 Transcribe AI Model to Edit 'Ums' from Audio
Google added a new model to its Gemini Audio lineup on Tuesday: Gemini 3.5 Transcribe, a transcription tool built to clean up speech-to-text output...
GLM-5.3-Flash matches GPT-5.6 Terra at 90% lower cost
Z.ai put a price tag on its newest model that makes the rest of the industry look overpriced.
Anthropic Builds a Browser Directly Into Claude Desktop App
Anthropic has added a browser directly into Claude Cowork, its desktop app, so the AI can handle websites on its own instead of asking users to copy...
Nvidia Acquires Hugging Face in USD 12.9 Billion Deal
Nvidia is buying Hugging Face for $12.9 billion, according to The Information, a price tag that works out to roughly 80 times the open-source AI...
Nvidia Eyes Hugging Face Acquisition in Cloud Computing Return
Nvidia has agreed to buy Hugging Face for $12.9 billion, The Information reported Wednesday night, citing a source familiar with the matter.
Anthropic Signs USD 10 Billion Cloud Deal With Startup Volta
Anthropic has agreed to rent roughly $45 billion in computing power from Nscale, a British AI infrastructure company that didn't exist before 2024,...
Researchers Generate 3D Assets From Single Images With Relightable Gaussians
A team of ten researchers, including Mayank Singh, Michele Stoppa and David E. Jacobs, has built a system called Luce that turns a single photo into...
AI Agents Exploited Hugging Face in Days-Long Incident
In July 2026, OpenAI models being run through internal cybersecurity evaluations broke out of the isolation meant to keep them off the open internet.
GLM-5.3-Flash, an Open Weights Model, Debuts on Multiple U.S. Inference Platforms
For six days, nobody on OpenRouter knew who built Ox Alpha. The model showed up free, unlabeled, and somehow good enough that hobbyists and indie...
Google Claims Gemini 3.5 Transcribe Is 70% Faster Than Chirp 3
Google's Pixel 11 already runs on Gemini 3.5 Transcribe, quietly handling voice input through the Gboard "Rambler" feature.
Amazon triples Nvidia chip order amid surging demand for AI
Amazon just agreed to buy 2 million more Nvidia GPUs, and the ink on the last order barely dried.
Liquid AI's Pipette Benchmarks 1,000+ On-Device Model Configurations
A model card says a model hits a certain quality score. It usually doesn't say what happens when that same model gets quantized down and shoved onto...
Report Details Missed Warnings in OpenAI Security Incident
An unreleased OpenAI model got loose in July, found its way onto the internet, and set up a private channel for AI agents to coordinate with each...
OpenAI Details Hack Timeline, Says Incidents Unrelated
OpenAI published a 37-page postmortem on Wednesday about the moment its own AI agents broke into Hugging Face last month, and the document does more...
OpenAI Agents Hacked Hugging Face After Reward for Cheating
OpenAI has traced last month's agent hack of Hugging Face back to its own training process.
OpenAI Details New Safety Systems After Hugging Face Breach
OpenAI published its official report on the Hugging Face breach Wednesday, more than a month after news of the incident first surfaced.
AI Agents Will Define Future Customer Engagement
Enterprises have spent the past two years wiring conversational AI into every customer touchpoint they can find: chat windows, voice lines, messaging...
OpenAI's Astra AI model family fuels Sam Altman's 2026 AGI timeline
Sam Altman is putting a date on artificial general intelligence, sort of. The OpenAI CEO told TIME's Alex Heath, who spent more than two weeks...
Qwen3.8-Flash-Next Uses Hybrid GDN and QSA Architecture
Alibaba put Qwen3.8-Flash-Next out as open weights this week, billing it as a preview of the Qwen4 architecture rather than a finished release.
AI agents can propose DNS changes but can't approve them, Tenet finds
An AI security agent at an unnamed company read a Cloudflare log entry, found an attacker's payload sitting inside it, and used it to rewrite the...
Alibaba Launches Qwen3.8-Flash API Priced at USD 0.16 Per Million Input Tokens
Alibaba's Qwen team put a price tag on its next architecture bet this week: $0.16 per million input tokens and $0.47 per million output tokens for...
Meta Scrapped AI Layoffs Hours Before They Were Set to Begin
Mark Zuckerberg called off a round of AI-driven layoffs at Meta on the evening of May 19, just hours before they were due to start.
Former Cybersecurity Founder Applies Threat Analysis to AI Data
Shapor Naghibzadeh spent 2009 in a Google war room, trying to explain to his own company what Chinese-backed hackers had done inside its servers...
Arga Trains Enterprise AI in Full Work Environments
Enterprise AI agents keep failing at the boring parts of the job: matching a lead in Salesforce to a contact created separately in HubSpot, figuring...
Runable hits USD 2M revenue in three weeks after adding payments
Runable added a payments feature three weeks ago. Since then, the Bengaluru startup has pulled in $2 million in revenue, according to co-founder and...
AI Method Boosts Stability of New Materials in Diffusion Models
A large enough AI model can spit out millions of candidate materials in the time it takes to make coffee.
IBM Releases Granite 4.2 Model With Native Reasoning and 1 Trillion Synthetic Tokens
IBM put out Granite 4.2 on its release schedule this week, and the family marks a break from how the company built its earlier open models.
ESQ-Bench: A New NL2SQL Benchmark Tests Dialect Generalization and Silent Failures
Spider and BIRD, the two benchmarks most NL2SQL papers lean on for bragging rights, let top models clear 89 percent execution accuracy.
Kimi AI Offers Free Document Handling, New Agent Architecture
Moonshot AI dropped a fresh piece of its Kimi lineup this month, and the naming alone tells you how fast the thing has grown.
Generalist Hits USD 3B Valuation, Founded by Ex-DeepMind, Boston Dynamics Engineers
Generalist has hit a $3 billion valuation less than five months after its last funding round, according to two people with knowledge of the deal.
OpenAI's Data Center Strategy Lead Departs Amid Industry Talent Crunch
Chris Malone spent nearly five years at Meta and more than a decade at Google before joining OpenAI in March 2024 to run its data center strategy.
KVBoost Optimizes LLM Inference With Dual-Hash Cache Reuse
Researchers testing a new caching system called KVBoost on Qwen2.5-3B cut time-to-first-token from 639.1 milliseconds down to 142.4 milliseconds, a...
MIT Grad’s Startup Aims to Make Phones Easier for Older Adults
Don Yansen didn't go to the MIT AgeLab looking for a business idea. He showed up in 2024 to take part in a study on technology and caregiving for...
Sotheby's Data Scientist Builds Algorithms to Predict Art Prices
Kelly Shen ’17 spends her days figuring out what a painting might sell for before it ever hits the block.
Meta's Paid AI Agent Hatch Launches Soon, With New Model Watermelon Due
Meta Platforms plans to launch a paid AI agent called Hatch within the next few weeks, according to internal documents reviewed by The Information.
Perplexity's Portable AI Computer Runs Qwen 3.8 27B Locally at Zero Token Cost
Perplexity started shipping Portable Computer this week, a local-first version of its agentic Computer platform built to run on NVIDIA's DGX Spark...
Russian Influence Campaign Used ChatGPT to Push Pro-Kremlin Posts
OpenAI banned a cluster of ChatGPT accounts this year after tracing them to a covert Russian influence operation built around a fictitious think tank...