AI News
AI news that matters. Updated daily.
No stories match your filters.
Musk vs. Altman Goes to Trial: What the Jury Actually Has to Decide
Three years after Elon Musk left OpenAI's board and two years after he filed suit, the biggest legal fight in AI history is finally in front of a jury. The case is not, despite how it gets covered, a philosophical debate about AI safety. It's a contract dispute - and the jury's job is to answer some very specific factual questions.
Ontario Audit: AI Medical Scribes Are Getting Basic Facts Wrong in Patient Notes
Ontario's healthcare auditors reviewed the AI note-taking tools that doctors use to document patient appointments - and found the tools are getting basic facts wrong, routinely.
Musk v. Altman Closing Arguments: A Bad Day for Musk's Legal Team
Closing arguments in the Musk v. Altman trial wrapped on May 14, and Elon Musk's legal team had a rough afternoon in court.
50+ Engineers Have Left SpaceXAI Since Its February Merger
More than 50 employees have reportedly walked out of SpaceXAI since February - a notable attrition rate for a company trying to compete at the frontier of AI model development.
Anthropic Open-Sources a Full Legal Plugin Suite for Claude
3,530 GitHub stars in under a month is a signal worth paying attention to. Anthropic's claude-for-legal repository, published April 21, is a free, open-source plugin suite that puts more than 50 named legal workflow agents into Claude - covering everything from NDA triage and DSAR responses to M&A diligence grids and patent claim charts.
OpenAI Codex Lands on iOS and Android in Preview for All ChatGPT Subscribers
Three months after Anthropic shipped remote monitoring for Claudee Code(/tools/claude-code/), OpenAI is adding the same capability to Codex - now available on iOS and Android in preview for all ChatGPT subscribers.
Anthropic Tops OpenAI in Business AI Adoption for the First Time
Three and a half years after ChatGPT made OpenAI the default answer to "what AI should our company use," businesses are choosing differently. For the first time, Anthropic has surpassed OpenAI in enterprise AI adoption, according to new market data tracking which AI platforms business teams are actually deploying.
NVIDIA RTX 5090 Price Hike Coming as GDDR7 Memory Costs Climb
NVIDIA is reportedly planning to raise prices on the RTX 5090, its flagship consumer GPU, with the broader RTX 50 series and PRO lineup potentially getting the same treatment. The culprit is GDDR7 memory - the high-bandwidth RAM soldered onto the card that makes it fast enough to run large AI models locally. GDDR7 costs have been climbing, and NVIDIA appears to be preparing to pass that along.
Richard Socher Raises $650M to Build AI That Researches and Improves Itself
$650 million. That's what Richard Socher - the AI researcher who led Salesforce's AI division and later built the search startup You.com - has raised to pursue one of the most ambitious ideas in AI: a system that can research and improve itself without human intervention.
Anthropic's 2028 AI Scenario Paper Is a Geopolitical Briefing, Not a Safety Study
The US leads China in frontier AI development today. Whether that's still true in 2028 depends on a handful of chip factories, some export regulations, and decisions being made right now - and Anthropic just published a paper making exactly that case.
OpenAI Enlists Outside Law Firm to Weigh Legal Action Against Apple
OpenAI has hired an outside law firm to evaluate legal options against Apple, according to Bloomberg. The nature of the dispute hasn't been disclosed publicly, but the move signals serious tension between two companies that announced a high-profile integration deal at WWDC 2024 - one where Apple would hand off certain Siri queries to ChatGPT rather than answering them natively through Apple Intelligence.
Clawdmeter Puts Claude Code Token Costs on Your Desktop in Real Time
A small open source project called Clawdmeter puts your Claudee Code(/tools/claude-code/) usage data on your desktop as a real-time widget. It's a menu bar or taskbar gadget that shows token consumption, session costs, and usage trends - the numbers Claude Code tracks locally but doesn't surface anywhere visible while you're working.
Musk-Altman Trial and Trump's China Trip: What Both Mean for AI Tools
Two storylines are advancing simultaneously this week that could reshape the conditions under which major AI companies operate: the Musk v. Altman trial and Trump's diplomatic engagement with China.
Microsoft Cancels Claude Code Licenses From Its Internal Developer Program
Five months after opening Claudee Code(/tools/claude-code/) to thousands of its own employees, Microsoft is pulling those licenses.
IBM Releases Open-Source Multilingual Embedding Models with 32K Context
IBM released Granite Embedding Multilingual R2 on Hugging Face today - a set of open-source embedding models available under the Apache 2.0 license.
Relying on AI for Everything Is Quietly Eroding Your Problem-Solving Fluency
What happens when you outsource thinking for months at a time? A developer named J. Pain documented the answer after leaning hard on AI tools: when he sat down to work without them, the process felt sluggish and unfamiliar. The mental fluency was gone.
Interactive Map Pinpoints AI Data Centers Across the US
When Oregon resident Isabelle Reksopuro heard Google was acquiring public land for data centers near The Dalles - a small city about 80 miles east of Portland - she couldn't sort out what was actually true. "There's a lot of misinformation about data centers," she told The Verge. "Google has denied taking that land." The experience pointed to a basic problem: AI infrastructure is everywhere, and almost nobody knows where it actually is.
Anthropic Deprecates Fixed Thinking Budget for Claude Opus 4.6 and Sonnet 4.6
Anthropic is removing the fixed-thinking-budget option for Claude Opus 4.6 and Sonnet 4.6. Developers who set a specific token limit for the model's reasoning process will need to switch to adaptive mode, where Claude decides how much to think based on the problem at hand.
VS Code's Local AI Model Support Still Requires a GitHub Copilot Plan
The promise of running AI coding assistants locally - no subscription fees, no data leaving your machine, no usage caps - just got more complicated. VS Code's new Agents window does let you point the editor at a locally-running AI model, but Microsoft kept the GitHubb Copilot(/tools/github-copilot/) subscription requirement in place. You still need an active Copilot plan and an internet connection to use it.
Cerebras IPO Raises $5.5B, Stock Doubles on First Day of Trading
108%. That's how much Cerebras stock surged on its first day of public trading on May 14, making it the first major tech IPO of 2026 and one of the more dramatic market debuts in years. The company raised $5.5 billion in the offering.
Gallup: 70% of Americans Oppose AI Data Centers in Their Communities
70 percent. That's how many Americans oppose AI data center construction in their communities, according to a new Gallup survey. Only 7 percent said they "strongly" support new facilities. And the survey's sharpest finding: Americans would rather live near a nuclear power plant than a data center.
Bench Founder Ian Crosby Gets $10M From Khosla for AI Bookkeeping Startup Synthetic
In December 2024, Ian Crosby's bookkeeping startup Bench collapsed almost overnight - locking hundreds of small business owners out of their financial records days before year-end tax deadlines, with almost no warning. Khosla Ventures has now bet $10 million that Crosby can build something better.
Claude Cracked an 11-Year-Old Bitcoin Wallet by Testing 3.5 Trillion Passwords
A Bitcoin trader just got $400,000 back from a wallet he'd written off more than a decade ago, and he did it with Claude.
Cisco Cuts Nearly 4,000 Jobs While Reporting Record Revenue, Cites AI Spending
Record quarterly revenue and nearly 4,000 fewer employees. That's Cisco's summary for its latest quarter - and a combination the company is framing explicitly as an AI bet.
Wirestock Raises $23M to Sell Creative Training Data to AI Labs
Three years ago, Wirestock was a marketplace where photographers and illustrators sold stock images. Today it's a training data supplier - and it just raised $23 million to keep building that business.
TechCrunch Startup Battlefield 200 Applications Close May 27
TechCrunch's Startup Battlefield 200 competition closes its application window on May 27. The prize: $100,000 in equity-free funding for one winner, plus editorial coverage from TechCrunch and direct exposure to the venture capital investors who attend TechCrunch Disrupt.
Anthropic Launches Claude Variant Aimed at Small Businesses
Anthropic released a new Claude variant on May 14 targeting small businesses - the company's most direct move yet toward the SMB market after years of focusing primarily on enterprise contracts and developer API access.
Claude Is Banning Paid Accounts Within Seconds of Signup
Some users are reporting a frustrating pattern: they create a Claude account, enter payment details, the charge goes through, and then they're immediately locked out with a ToS violation notice — all within the same minute.
OpenAI Brings Codex Coding Agent to ChatGPT Mobile App
OpenAI's Codex coding agent - the feature inside ChatGPT that can write, run, and debug code autonomously across a codebase - is now accessible from the ChatGPT mobile app. The update, announced May 14, lets you monitor active tasks, redirect them if they drift off course, and approve next steps, all from your phone.
NVIDIA Releases Quantized Kimi-K2.6 and Kimi-K2.5 for Local GPU Deployment
NVIDIA has published NVFP4-quantized versions of Moonshot AI's Kimi-K2.6 and Kimi-K2.5 models, both cleared for commercial use on local NVIDIA hardware. The quantization was performed using NVIDIA's Model Optimizer.
AI Watch Designs Went Viral. Chinese Factories Are Now Making Them.
What happens when an AI-generated product concept goes viral before anyone checks whether it actually exists?
Claude Helped Someone Crack Open a $400K Bitcoin Wallet Locked for 11 Years
$400,000. That's how much Bitcoin sat locked in a forgotten wallet for 11 years — until Claude helped recover it.
Personal Software Is Becoming Real: Build the Exact App You Need Without Hiring a Developer
The dream of software that does exactly what you need has been theoretical for most of computing history. You got the features some product team decided to build, in the interface they designed. Want something else? Learn to code - or accept the compromise.
Gen Z's Blurry Line Between Fact and Feeling Is Now an AI Content Problem
Something specific is shifting in how younger users evaluate information online, and it predates AI tools by nearly a decade. What AI tools are doing now is accelerating the problem for everyone.
AWS User Hits $30,000 Bedrock Bill as Cost Anomaly Detection Fails
$30,000. That's the invoice one AWS customer received after a Claude session on Amazon Bedrock ran unchecked - and the billing protection tool AWS markets as the safety net for exactly this scenario did nothing to stop it.
Meta Cuts 10% of Staff as Profits Rise and Morale Falls
Record profits. Falling morale. Next week, Meta begins cutting roughly 10% of its workforce - tens of thousands of people at a company with around 74,000 employees globally.
Local LLMs as Personal Knowledge Bases: The Setup, the Friction, and the Payoff
The quiet use case in local AI communities right now is not about coding. It's about something more personal: running your own notes, PDFs, and documents through a local language model to query your own life - without any of it ever leaving your machine.
Veo 3's 3-Video Daily Cap Is Killing Professional Workflows
Three videos per day. That's Google's current hard cap for Veo 3 generation through Gemini Advanced, and for anyone running iterative video production, it's effectively useless.
Agentic AI Workflows Are Breaking the Flat-Rate Subscription Model
There's a quiet tension building between how AI subscriptions were priced and how people are actually using them. When Claude, ChatGPT, and similar tools launched their $20/month tiers, the assumption was simple: one human, typing questions, reading answers, a few hours a day. That model held for about 18 months.
Why AI Writes With Em-Dashes: The Books Hidden in Your Chatbot's Training Data
Ask a thousand people to identify AI-generated text and most will point to the em-dashes. ChatGPT and Claude write like copy editors who read too many literary magazines - which is odd, because the common assumption is that these models learned to write from the internet.
The Gap Between AI Companies and Their Users That Nobody Is Closing
"The conversation is sort of happening in Silicon Valley around one thing, and a totally different conversation is happening among consumers." That's Campbell Brown, former head of news partnerships at Meta, speaking at StrictlyVC this week. The disconnect she's naming is real, it's widening, and it has practical consequences for anyone who depends on AI tools.
Clio Hits $500M ARR as Legal Tech's Long-Delayed Adoption Wave Finally Arrives
$500 million in annual recurring revenue. Clio just crossed that mark, putting the legal practice management platform in rare company for a vertical SaaS business and confirming what adoption numbers have been suggesting for months: law firms are finally buying software.
AI Tooling Complexity Is Growing Faster Than the Tools Themselves
Last year, the pitch was simple: call an API, get a useful result, ship something. Now, building even a basic AI-powered product means navigating vector databases (external storage systems that let AI search through your data by meaning rather than keywords), orchestration frameworks (tools that chain multiple AI steps together), embedding pipelines, context window management, and a half-dozen competing standards for how agents should call tools.
Community Dashboard Tracks How AI Model ELO Ratings Change After Launch
If you've ever felt like ChatGPT or Claude got noticeably worse a few weeks after a new version launched, you're not imagining it. A developer has built a dashboard that visualizes historical ELO ratings from Chatbot Arena - the crowd-sourced benchmark where real users vote on which AI response they prefer - to test exactly that hypothesis.
Open-Source Plugin Adds Learning Exercises to Claude Code and Codex
Most developers using AI coding assistants have noticed the pattern: Claudee Code(/tools/claude-code/) or Codex writes the code, you accept it, ship it, and three weeks later you can't explain how any of it works. A new open-source plugin directly addresses this.
AI Referral Traffic Is Tiny But Converts Better Than Google. Here's the Data.
210 sessions in 90 days. That sounds like a rounding error in a standard analytics dashboard. But when 70% of those visitors are actively engaging with content — clicking, scrolling, spending time — the number starts looking different.
Claude Code Hook Lets You Run 24/7 Automation Without -p Flag or SDK Credits
Anthropic is ending subsidized usage of Claudee Code(/tools/claude-code/)'s -p flag on June 15, 2026. The flag lets users run Claude Code in non-interactive mode - single prompt in, single response out, no persistent session - which has become a common pattern for scripting and lightweight automation. Once the change hits, that usage will draw from API credits rather than the standard subscription.
Anthropic Forms $200M Partnership With the Gates Foundation
Anthropic has formed a $200 million partnership with the Bill & Melinda Gates Foundation, directing Claude and the company's AI research toward global health and development challenges.
ChatGPT Now Tracks Mental Health Red Flags Across Full Conversations
Single messages rarely tell the whole story. Someone asking about medication dosages once isn't automatically alarming - but asking repeatedly, alongside expressions of hopelessness, across a long conversation, is a different pattern entirely. OpenAI's new safety update trains ChatGPT to make that distinction.
How Async Continuous Batching Reduces Latency in AI Inference Servers
Continuous batching made AI inference servers dramatically more efficient when the technique arrived a few years ago. A new approach described by Hugging Face takes it further by making the process asynchronous - and the difference is measurable for anyone running AI at meaningful request volume.