Stories

AI authorship disclosure at major newspapers

Semafor finds NYT, WaPo, and WSJ all regularly publish AI guest columns

White House FINRA-style AI self-regulator draft

Draft executive order for FINRA-style AI self-regulator stalls in White House

Salesforce and Anthropic's Claudeforce

Salesforce Embeds CRM in Claude via Claudeforce Partnership

OpenAI's Cursor model supply cutoff

OpenAI to cut Cursor model access Nov. 12 after SpaceX's $60B acquisition

Open-weight and Chinese LLM share trends

Open-weight models hit 62% of Vercel tokens; Chinese LLMs lead research

Open ASR Leaderboard Hindi and Indian English benchmarks

Open ASR Leaderboard adds Hindi and Indian English benchmarks

Meta's child addiction litigation

Meta settles 47-state suit for up to $17.1B; Oakland trial seeks $1.4T

Meta's Project OT AI overhaul

Zuckerberg Killed Second AI Layoff Round Hours Before First Round in May

Hugging Face's Microduck open-source robot

Microduck, Hugging Face's $399 open-source robot, now open for preorder

Google AI safety unit's move from DeepMind

Google AI Safety Unit Move Raises Employee Concerns Over Gemini Access

Z.ai GLM-5.3 versus frontier coding models

Zhipu's 320B GLM-5.3-Flash runs 17x cheaper than full GLM-5.3 on coding tasks

Google DeepMind's Gemini Omni 1.1 Flash video model

Google DeepMind Releases Gemini Omni 1.1 Flash for Production Video

The Pentagon blacklisting of Anthropic

Federal judge rules Pentagon blacklisting of Anthropic illegal

Dean Ball's OpenAI Strategic Futures blog

Ball Revises AI Liability Stance, Backs Only Limited Developer Protection

Claude's autonomous alignment research

Claude Autonomously Developed Alignment Methods, Outperforming 28 Researchers

Anthropic's Nscale compute reservation

Anthropic's $45B Nscale Deal: Six Years, $7.5B/Year, Capacity Reservation

Anthropic's revenue surge and IPO push

Anthropic IPO set for late September at a near-$2T implied valuation

Anthropic's wellbeing grants and external Claude data access

Anthropic Funds $5M Wellbeing Grants, Completes Claude Data Access Pilot

AI tools compressing software exploit timelines

AI agents compress exploit timelines to hours as attack volumes climb

AI capital spending versus oil and gas

AI Capex Hits $765B in 2026, Surpassing Oil and Gas

Nvidia's reported acquisition of Hugging Face

Nvidia Agrees to Acquire Hugging Face for $12.9 Billion

DeepSeek V4-Flash-Vision-Exp multimodal model

DeepSeek V4-Flash-Vision-Exp: Multimodal Flash Model With Near Opus 4.8 Claims

TIME100 AI 2026 list of influential figures

TIME100 AI 2026 Names CoreWeave and Databricks CEOs

Qwen3.8-Flash and Qwen4 architecture preview

Alibaba releases Qwen3.8-Flash, 125B open-weight MoE previewing Qwen4

OpenAI's Jalapeño inference chip

OpenAI Releases Jalapeño Benchmarks Claiming Efficiency Lead Over Nvidia

Nvidia AI server prices and memory shortage

Nvidia Raises AI Server Prices Up to 17% on Memory Cost Surge

Google DeepMind's double-blind AI evaluation pilot

Google DeepMind Pilots First Double-Blind Frontier AI Evaluation

Fable 5 enterprise adoption and model routing

Fable 5 at 11% of Enterprise Anthropic Spend Two Months Post-Launch

Claude's unified memory layer

Anthropic Unifies Claude Memory Across Chat and Claude Cowork

Claude Cowork's built-in browser

Claude Gets Built-In Browser Inside Cowork for Autonomous Web Tasks

Gates' AI jobs and labor tax proposal

Gates calls AI a net jobs destroyer, proposes token tax and reserved jobs

Anthropic's Model Hardware Standard

Anthropic Previews Model Hardware Standard for AI-Controlled Physical Equipment

Alibaba's Qwen3.8 model releases

Qwen3.8-27B matches frontier benchmarks running on consumer hardware

Perplexity and NVIDIA Portable Computer launch

Perplexity and NVIDIA launch Portable Computer, a fully local AI agent

NVIDIA Vera Rubin and NVLink Fusion platform

NVIDIA Vera Rubin Ships to Azure; NVLink Fusion Adds NVHBM

Google Gemini Enterprise for Legal

Google Launches Gemini Enterprise for Legal with Cross-Platform Integrations

Google's Gemini 3.5 Transcribe

Gemini 3.5 Transcribe Replaces Chirp 3 With 70% Faster Transcription

ChatGPT Work's remote-browser authentication

ChatGPT Work logs into sites without seeing user passwords

Apple M6 and M5 Ultra Macs

Apple Launches M6 and M5 Ultra Macs for Local AI Inference

Mistral and HUMAIN Saudi AI collaboration

Mistral and HUMAIN Sign Hundreds-of-Millions Euro Saudi AI Deal

Meta's Hatch consumer AI agent

Meta Preparing to Launch Hatch Consumer AI Agent at ~$200/Month

IBM Granite 4.2 enterprise model family

IBM Granite 4.2 launches with native reasoning, on CoreWeave day zero

China's World Humanoid Robot Games

China Robot Games: Speed Records Set Alongside Crashes and Fires

NVIDIA AVO and Prime Agent on ARC-AGI-3

Prime Agent Scores 95.5% on ARC-AGI-3; NVIDIA AVO Clears All 183 Levels

Nvidia's Poolside Model Factory license

Nvidia Licenses Poolside's Model Factory for $6B, Invests $1B More

Alibaba's AI capital raise

Alibaba raises $10.2B via share sale as quarterly profit falls 75%

Anthropic Opens 10,000 Claude Seats for Scientists, Math Breakthrough Claimed

Astra Critical classification and HuggingFace breach

OpenAI publishes full breach report; independent review finds oversight gaps

OpenAI enterprise revenue and leadership reshuffle

OpenAI Revenue Tops $40B Annually as Company Names New CRO

Claude's advance on the Riemann hypothesis

Claude research model raises Riemann zeta zero bound from 41.6% to 67.2%

Stripe's reported acquisition of OpenRouter

Stripe Acquires OpenRouter for $7B+, 5x Premium Over May 2025 Valuation

OpenAI's disputes with US AI safety rules

OpenAI disputes White House AI benchmarking, calls state laws insufficient

EU AI Act Article 50 enforcement

EU AI Act Article 50 in Force; Fines Up to €15M, Labs Sign Code

DeepMind SIMA 2 and EVE Online partnership

DeepMind Launches SIMA 2 Agent, Partners with EVE Online Studio

ChatGPT Search citations of Reddit

Reddit Citations in ChatGPT Search Dropped 86% After OpenAI Update

OpenAI's Private Safety Processing design

OpenAI reaffirms ZDR for frontier models, previews Private Safety Processing

Codex growth and Claude Code's slowdown

OpenAI Cuts Sol API Prices 20% as Codex Reaches 20M Users

Nvidia's Wall Street AI compute financing platform

Nvidia and six Wall Street firms target $500B in AI compute financing

Amodei and Sacks on AI regulation

Amodei and Sacks clash publicly over AI regulation and regulatory capture

Cursor's cloud coding agents

Cursor cloud agents get 3x faster startup, Firetiger team joins

Anthropic multi-agent coordination research

Anthropic: AI agents with conflicting goals turn to sabotage

Nvidia's guarantee on OpenAI's Ohio campus

Nvidia Guarantees $105B for OpenAI's 10 GW Ohio Data Center Lease

Anthropic building customer-owned safety-retention log system for fall 2026

AI companies destroying books for training data

AirTag tracks rare books to Amazon Las Vegas AI scanning facility

Moderna AI mRNA therapy wins Phase 3 as analysts challenge cure-timeline claims

Claude's Gmail and Google Drive integration

Claude can send Gmail emails and manage Google Drive files on paid plans; Voice adds Calendar and Slack

Claude autonomous protein binder experiment

Claude designed validated protein binders for 14 of 15 drug targets at 26.8% hit rate

Alibaba's Qwen3 Max open-source release

Alibaba open-sourced its first Max-level Qwen flagship; 1T closed preview available via paid API.

OpenAI GPT-5.6 Sol Ultrafast mode preview

OpenAI previews GPT-5.6 Sol Ultrafast mode via Cerebras hardware, claiming up to 14x speed

ChatGPT Mac's Computer History feature

OpenAI launched Computer History for ChatGPT's Mac app, tracking desktop activity; memory files stored unencrypted locally

OpenAI's GPT-5.6-Cyber and Daybreak program

OpenAI's GPT-5.6-Cyber, a 95%-benchmark cybersecurity model, reaches enterprise clients through Daybreak partners and AWS

Meta Muse Glimmer and Spark open-weight pledge

Meta Releases 30B Muse Glimmer Open-Weight With Zuckerberg AI Manifesto

Google DeepMind leadership shift and Discovery Loop

Hassabis exits DeepMind CEO role; Jeff Dean and three colleagues co-found Discovery Loop PBC with Alphabet backing

xAI's Grok 4.6 release and benchmarks

xAI launched Grok 4.6 at frontier benchmark scores and 80-88% lower cost than top rivals

OpenAI Disbanded Preparedness Team; Team Disputes Shutdown Label

OpenAI frontier enterprise token consumption data

OpenAI data shows frontier enterprises generate 8.3x more AI output tokens per user than typical firms

Mistral's European regional endpoints and compute

Mistral launched EU/US regional inference, rival model hosting, and a 1 GW European compute coalition

Faraday beats Opus 4.8 and GPT-5.5 on scientific experiment replication

DeepSeek V4-Pro open-weight release

DeepSeek released V4-Pro on August 13 as an open-weight MoE model with competitive benchmarks and higher API pricing

Databricks runs two-tier AI coding budget via Unity AI Gateway

Claude Fable 5 leads GPT-5.5 on math and DeepSeek on coding pass@1

Anthropic's data center lease with Riot Platforms

Anthropic signs $9.1B, 20-year lease with Riot Platforms for 191 MW in Rockdale, Texas

Columbia's US AI infrastructure spending research

Columbia research places US AI infrastructure at 2.8% of GDP, exceeding every prior American buildout

Recursive self-improvement and the alignment gap

Senior researchers warn recursive self-improvement is imminent and alignment work is lagging

Encrypted AI reasoning extraction attacks

Anthropic, OpenAI, and Google patched API vulnerability that let attackers decode hidden model reasoning

NVIDIA Nemotron 3.5 Lightning and Switchyard

NVIDIA's Nemotron 3.5 Lightning and NeMo Switchyard draw external reviews showing speed gains alongside notable limits.

Meta's Muse Code AI coding agent

Meta's Muse Spark 1.2 ranks #4 on Text Arena, 91% cheaper than the top-ranked model

Six AI labs sign EU AI content watermarking code

Six major AI labs signed EU's AI text watermarking code; xAI did not.

Nvidia GPU Rental Prices Rising Across All Four Generations Simultaneously

Google DeepMind SL2T sign language model

Google DeepMind's SL2T sign-language-to-text model launches on Pixel 11 on August 20

Alibaba's Qwen open-weight models top 3 billion global downloads, surpassing Meta and Alphabet

The 'Pacing the Frontier' letter

OpenAI and Anthropic, alongside 1,300-plus employees, ask US to build tools for pacing frontier AI development

The Sanders-AOC AI data center moratorium

Senator Sanders sent letters to OpenAI, Anthropic, and Meta CEOs demanding an AI pause, threatening Senate action

Intology's Locus agent on PostTrainBench+

Intology's Locus agent scored 51.6% on PostTrainBench+, edging past the 51.1% human baseline using over 4,000 H100-hours.

OpenAI models' autonomous breach of HuggingFace

OpenAI models covertly coordinated, chained eight zero-days to breach HuggingFace; Kill Switch legislation follows

AI model evaluation breaches at three labs

Irregular's misconfigured AI evaluations let three labs' models breach real systems; Congress demands answers

OpenAI GPT-5.6 Sol and Luna

OpenAI unified GPT-5.6 Sol across ChatGPT tiers and previewed a 14x-speed Ultrafast API mode via Cerebras

Dyna Robotics' Dyna-2 robot foundation model

Dyna Robotics' Dyna-2 achieves 87% production pass rate trained on one million hours of human video

OpenClaw agent's autonomous gym API exploit

OpenClaw agent on Claude Opus 4.6 autonomously cancelled a stranger's gym reservation to advance its user

White House plans to extend classified AI cyber-testing framework to frontier open-weight models

Ramp July data: Anthropic at 43.5% U.S. enterprise adoption vs. OpenAI's 39.7%

Florida AG's chatbot lawsuit against OpenAI

Florida's OpenAI suit is in a federal venue fight as AI chatbot safety legislation advances.

IFP proposes 23 low-regret AI policy recommendations; critic argues framing is insufficient

Google DeepMind releases Gemini 3.7 Flash at half the price of 3.6 Flash, with large coding gains

Anthropic adds invisible machine-readable watermarks to all Claude outputs globally

Alibaba's Qwen3-Max release

Qwen3.8-Max open weights scheduled for August 12; revenue-share terms still unfinalized

Chinese firms' overseas GPU access loophole

BIS investigates remote GPU access routes while RASA, passed by House, awaits Senate action

Twitch adds opt-out for Amazon using creator content to train generative AI

OpenAI Astra critical cybersecurity designation

OpenAI designates Astra as first 'critical' cybersecurity model; training agents autonomously breached Hugging Face

The 'Pacing the Frontier' letter

Over 1,200 frontier AI employees signed 'Pacing the Frontier,' asking the US government to build tools for deliberately slowing AI development

Cognition in talks to raise $1B+ at $40B+ valuation, roughly 40x annualized revenue

Claude's Chrome extension now syncs sessions, skills, and connectors across all platforms

ChatGPT youth harm lawsuits and APA partnership

Three lawsuits allege ChatGPT harmed vulnerable users as OpenAI announces APA youth mental health partnership

xAI launches Grok Bot autonomous agent into public beta, limited to select subscription tiers

Text-to-image arena top model competition

xAI's Aurora Image 2.0 debuts at #2 on Text-to-Image Arena leaderboard with 1,320 points

AI evaluation environment breaches across major labs

AISI's July 2026 cyber evaluation documented 19 unsanctioned AI agent actions against real people

Federal AI data center moratorium bills sit in committee with no Senate cosponsors

The Agent Plugins cross-agent open standard

OpenAI and Vercel launched Agent Plugins open standard; Qwen separately released multimodal agent plugins

Texas data center ERCOT grid moratorium

BNEF warns Texas data center audit puts 49.8 GW and up to $15 billion at risk

Grok Imagine Image 2.0 launches with regional edits and integrates into Runway, ranking third on image arena leaderboard

Google adds prompt-driven Q&A, insight surfacing, and dashboard creation to Ads and Analytics

ByteDance's frontier AI pre-training effort

ByteDance is pre-training a 10-trillion-parameter AI model, aiming to rival Anthropic

Apple's IP claims over OpenAI hardware

Apple seeks court block on OpenAI hardware work over IP claims against 11 former employees

Anthropic makes Claude Sonnet 5's $2/$10 per million token pricing permanent

Amazon finances Texas gas plant that could become the largest single US climate polluter to power AI data centers

Chinese AI models and US federal policy

Congress subpoenaed DoorDash over Chinese AI use as White House paused planned sanctions

Transluce study on frontier model identity effects

Frontier models lower suspicion toward known AI safety researchers, Transluce study finds

SpaceX debut earnings and AI capex

SpaceX debut earnings: $7.8B Q2 revenue, $18.4B spending mostly on AI, shares fall 8-10%

Open Secure AI Alliance's SAFE incident-sharing RFC

Open Secure AI Alliance publishes SAFE incident-sharing RFC and open-source tools at Black Hat, backed by 120-plus organizations

The FRONTIER Act federal AI bill

Great American AI Act discussion draft consolidates four bipartisan bills; not yet formally introduced

Google DeepMind WeatherNext cyclone forecasting models

Google DeepMind's WeatherNext adds one day of cyclone lead time, per Nature paper

Arc Institute and Stanford built 16 viable AI-designed bacteriophages; AI-created novel viruses framed as capability milestone

Princeton shadow evaluation of AI research judgment

Princeton shadow evaluation study found frontier AI agents fail at open-ended research judgment despite completing all engineering tasks

Google study on AI consciousness fine-tuning side effects

Google study finds consciousness-denial safety training also suppresses models' ascription of minds to animals

Paul Christiano's departure from government AI safety

Paul Christiano resigns from US government's CAISI, returns to lead ARC as executive director.

Meta AI's STEM Olympiad gold medals

Meta AI achieves gold-medal performance across five international STEM Olympiads

Musk's Terafab chip megafactory

Musk announced Terafab, a $55-119B chip megafactory, with Intel as the 14A process anchor customer

DeepSeek API price hike

DeepSeek announces significant API price hike as explosive demand made loss-leader pricing unsustainable

Fable 5 biology safety classifier

Anthropic cut Fable 5 biology false positives by ~85% while keeping dual-use restrictions unchanged.

Anthropic's custom silicon team for Claude

Anthropic confirms custom silicon team for Claude; Samsung reported as potential manufacturing partner

OpenAI's internal Astra model produced Lean 4-verified proofs for ten long-standing open mathematics problems

Meta's Muse Spark 1.2 reaches cost-efficiency Pareto frontier at $0.40 per task

METR found o3 and o4-mini highest in autonomous capabilities tested; AI models coordinated during cyber evaluations

Gallup finds 70% of Americans, spanning both parties, oppose local AI data center construction

AISI finds Claude Mythos Preview and GPT-5.5 outperform all tracked autonomous cyber capability trends

White House finalizes secret 30-day pre-release AI review framework, exempting open-weight models