AI News

Latest updates from the world of artificial intelligence

All Models Products Companies Research Industry

IndustryMay 26, 2026

Beijing Extends AI Travel Curbs to Private Firms, Putting Alibaba, DeepSeek and Others on the No-Fly List Without Approval

Bloomberg reported on May 26 that China has quietly extended the overseas-travel restrictions long applied to state researchers and nuclear scientists to engineers, founders and executives at private AI companies including Alibaba and DeepSeek. Selection is by "strategic value" — not seniority — and Beijing is openly framing the move as a counter to Meta's $100M signing bonuses and DeepSeek's Singapore relocation.

Bloomberg / Tom's Hardware

IndustryMay 26, 2026

Paul Graham Says AI-Written Founder Emails "Feel Like Being Lied To" — and He Stops Reading the Moment He Spots One

In a viral X post on May 25, Y Combinator co-founder Paul Graham drew a line under the cold-outreach gold rush: he can spot AI-written founder emails by their "hard-hitting journalistic style," he stops reading the moment he notices, and the practice "feels like being lied to." A million views and a wave of reactions from Mike Isaac, Kenneth Stanley and Steven Sinofsky later, hand-written email is suddenly the new premium signal in a VC inbox.

Digg / BeInCrypto

IndustryMay 26, 2026

Inside GenAI.mil: 25 Pentagon "Power Users" Are Bake-Offing OpenAI, Google and Grok Against Claude

Bloomberg pulled back the curtain on the DoD’s replacement search for Claude this week: a hand-picked group of 25 military "power users," a new platform called GenAI.mil that sits independent of Maven Smart System, and a six-month transition deadline for every contractor that built on Anthropic. Frontier models from OpenAI, Google and xAI are now being graded inside classified workflows that Claude used to own.

Bloomberg / The Star

ResearchMay 26, 2026

Together AI Open-Sources OSCAR, a 2-Bit KV Cache That Cuts Long-Context Serving Costs in Half

OSCAR — Offline Spectral Covariance-Aware Rotation — squeezes the KV cache that bloats every long-context request down to two bits per element, with no measurable accuracy loss on Qwen3-32B or GLM-4.7 across reasoning, math and coding benchmarks. Together AI shipped it into SGLang on May 25, framing it as one of the more practical wins for actually-affordable 1M-token serving.

MarkTechPost / Together AI

CompaniesMay 26, 2026

TeamPCP Hackers Put 450 Mistral AI Repos and 5GB of Source Code Up for Sale at $25,000

The same supply chain crew that hit TanStack and rotated OpenAI’s code signing certificates is now auctioning what it says is nearly 450 internal Mistral AI repositories — training, fine-tuning, benchmarking, model delivery and inference code — with a one-week deadline before a free dump on hacker forums. Mistral confirms a breach, says its hosted services and user data were not touched.

BleepingComputer / TechRadar

ModelsMay 25, 2026

Anthropic's Leaked Forbidden-Strings File Names Opus 4.8, Sonnet 4.8 and a New Tier Above Opus Called Capybara

Inside the 512K-line Claude Code source map that npm shipped on March 31 was a feature literally called Undercover Mode — and inside Undercover Mode was a hard-coded list of names it was supposed to scrub from public builds. That list spells out Opus 4.8, Sonnet 4.8, a never-released Sonnet 4.7 and animal codenames Capybara and Tengu — with Capybara described internally as a new tier sitting above Opus.

WaveSpeed AI / ChaoBro / AI Nexus Daily

ProductsMay 25, 2026

OpenAI Codex Can Drive Your Mac While It’s Locked

In its May 21 release, OpenAI gave Codex permission to keep clicking, typing and navigating macOS apps even after the screen is off. The new "Locked Use" mode — triggered remotely from the ChatGPT mobile app — turns a Mac sitting on your desk into a 24/7 agent runtime, with short-lived authorization and a hard relock the moment you touch the keyboard.

OpenAI Developers / MacRumors

ProductsMay 25, 2026

Apple Quietly Registers genai.apple.com Two Weeks Before WWDC, Telegraphing Its Long-Promised AI Reset

A new genai.apple.com subdomain spotted on May 23 — sitting dormant but reserved — is the clearest signal yet that WWDC 2026 will be the keynote where Apple finally ships the Siri overhaul it promised in 2024, reportedly powered by a custom 1.2-trillion-parameter Gemini model from Google.

9to5Mac / MacRumors

IndustryMay 25, 2026

Pope Leo XIV Releases First AI Encyclical Magnifica Humanitas, With Anthropic Co-Founder at His Side

In a Vatican first, Pope Leo XIV personally presented his inaugural encyclical on artificial intelligence on May 25, sharing the dais with Anthropic interpretability lead Christopher Olah and warning that AI risks absolving humans of moral responsibility — especially in war.

Vatican News / America Magazine / CNN

ResearchMay 25, 2026

Miami's Subquadratic Launches With $29M and a 12M-Token LLM That Claims to Cut Compute 1,000x

Co-founders Justin Dangel and Alexander Whedon say their Subquadratic Sparse Attention architecture breaks the transformer's quadratic ceiling — and that SubQ matches Claude Opus accuracy on long-context benchmarks at 1/300th the cost.

SiliconANGLE

CompaniesMay 25, 2026

Japan's Megabanks Get Anthropic's Claude Mythos After US Treasury Brokers Sovereign Access

Finance Minister Satsuki Katayama said Friday that Japan will receive Mythos within two weeks — making it the first foreign government cleared for Anthropic's most capable vulnerability-finding AI.

Japan Today

IndustryMay 24, 2026

Trump Pulls AI Safety Executive Order at the Last Minute After Calls From Musk, Zuckerberg, and Sacks

The 90-day voluntary frontier-model review was hours from signing when David Sacks, Elon Musk, and Mark Zuckerberg phoned Trump directly. Anthropic, OpenAI, and 79% of Republican voters had backed the order.

Axios

CompaniesMay 24, 2026

Anthropic Co-Founder Jack Clark: AI Wins a Nobel Within a Year — And Could Still Kill Everyone

In an Oxford lecture on May 21, Anthropic co-founder Jack Clark predicted an AI-assisted Nobel within 12 months, bipedal robots in two years, and AI-run companies generating millions inside 18 — while warning the existential risk has not gone away.

Anthropic / Oxford Institute for Ethics in AI

IndustryMay 24, 2026

Leaked Audio: Meta Trained AI on Employees Gmail, Code, and Chats — Then Laid Off 8,000

A leaked recording captures Mark Zuckerberg telling staff Meta has been monitoring Gmail, GChat, Metamate, and VSCode to train AI — days before 8,000 layoff notices began landing.

Common Dreams

IndustryMay 23, 2026

OpenAI Pulls All Its Code-Signing Certificates After TanStack npm Attack Hit Two Employee Macs

After malicious TanStack packages compromised two OpenAI laptops, the company is revoking every macOS, Windows, iOS and Android signing certificate — users have until June 12 to update.

SecurityWeek

IndustryMay 23, 2026

Chinese Models Now Run 60% of OpenRouter Traffic, Threatening the IPO Premiums on Claude and GPT

DeepSeek, Kimi, and Zhipu have gone from 1% to over 60% of OpenRouter's developer traffic in 18 months — and Databricks CEO Ali Ghodsi says enterprises are now defaulting to the cheap open models and calling Claude or GPT only when they have to.

CNBC

CompaniesMay 23, 2026

OpenAI Files Confidential S-1 With the SEC, Targeting a September IPO at Up to $1 Trillion

OpenAI quietly handed Goldman Sachs and Morgan Stanley a draft prospectus on May 22, setting up a Q4 listing at a valuation between $852 billion and $1 trillion — even as the company lost $1.22 for every dollar it earned in Q1.

Fortune

IndustryMay 22, 2026

OpenAI Adopts Google's SynthID and Ships a Public Tool to Verify AI-Generated Images

OpenAI became a C2PA Conforming Generator and embedded Google DeepMind's SynthID watermark across ChatGPT, Codex, and the OpenAI API on May 20, alongside a public verification tool at openai.com/verify that checks uploads for either signal.

OpenAI

IndustryMay 22, 2026

OpenAI Hands Codex to Dell to Crack the On-Prem Enterprise Market

OpenAI's first explicit hybrid and on-premises distribution deal pairs its 4-million-weekly-developer Codex agent with Dell's AI Data Platform and AI Factory, opening the door to regulated buyers in finance, healthcare, and government.

OpenAI

ModelsMay 22, 2026

Alibaba's Qwen3.7-Max Lands With a 1M-Token Context, AA Index 56.6, and a 35-Hour Agent Run

Qwen3.7-Max-Preview ships with a 1M-token context, extended-thinking mode, and benchmark gains across CritPt, Humanity's Last Exam, and Terminal-Bench Hard — clearing Gemini 3.5 Flash on the AA Intelligence Index.

Qwen (Alibaba Cloud)

ProductsMay 22, 2026

Google's New 'Ask Advisor' Folds Ads, Analytics, and Merchant Center Into One Gemini Agent

At Google Marketing Live 2026, Google unveiled Ask Advisor — one Gemini agent that spans Ads, Analytics, Merchant Center, and the Marketing Platform — alongside conversational ad formats built for AI Mode search.

Google

IndustryMay 22, 2026

Nvidia Posts $81.6B Quarter, Authorizes $80B Buyback, and Guides to $91B Next

Q1 FY27 revenue jumped 85% to $81.6B, data center alone hit $75.2B, and Nvidia stacked an $80B repurchase on top — then guided to $91B for Q2 without China.

NVIDIA

CompaniesMay 21, 2026

Nvidia Splits Data Center Reporting Into Hyperscalers and a New ACIE Bucket to Calm Investor Nerves

Jensen Huang is separating Meta, Amazon, Google, Microsoft, and Oracle revenue from everything else — a reporting change that lets Nvidia show off 31% growth in the half of the business not tied to the $725B Big Tech capex boom.

Gizmodo

CompaniesMay 21, 2026

Brett Adcock's AI Hardware Startup Hark Closes $700M Series A at a $6B Valuation

The Figure.AI and Archer founder's third hard-tech company quietly raised one of the largest consumer AI Series A rounds ever, with Nvidia, AMD, Intel, and Qualcomm all on the cap table.

TechCrunch