WSJ reviews Sam Altman's 80+ personal investments, 10 of which pursued business with OpenAI, as Congress investigates ahead of OpenAI's planned IPO - The Wall Street Journal

OpenAI launches GPT-5.5-Cyber with 85.6% on CyberGym, expanding its Daybreak defensive security program. - OpenAI

OpenAI announced a broad cybersecurity initiative including a 'Patch The Planet' pledge and multiple vulnerability-prevention programs. - theregister

OpenAI Codex now supports up to eight parallel subagents with isolated context windows for long-running coding tasks. - 4sysops

odds of GPT-5.6 before June 30 have plummeted mysteriously on Kalshi

Eyes Wide Shut: CFR warns the Anthropic export ban is alienating Five Eyes allies and squandering America's three-to-eight-month lead over Chinese AI - Council on Foreign Relations

Anthropic's suspension of foreign access to its top models exposed European businesses' dependence on U.S. AI providers. - Fortune

Mythos impossible: Anthropic's Mythos AI reportedly penetrated nearly all NSA classified systems in hours during an authorized red-team exercise. - TechSpot

Alphabet shed up to $270 billion in market cap after Nobel laureate John Jumper left DeepMind for Anthropic and Gemini co-lead Noam Shazeer joined OpenAI. - New York Post

Seoul Crushing: SK Hynix slowdown report triggers 10% Kospi collapse, dragging Nasdaq 100 futures down 2.8% and European stocks lower - Bloomberg

Taiwan's stock market has doubled in a year on AI euphoria, but retail investors borrowing to buy are raising bubble fears - Bloomberg

Goldman Sachs warns AI stock gains are outpacing realistic revenue projections and recommends hedging with put options - MarketWatch

One does not leave a convivial party before closing time. - G. K. Chesterton

I come to bury Caesar, not to praise him: Michael Burry warns Nvidia faces aggressive stock fall due to highly concentrated customer base and financial indicators. - Fast Company

Google DeepMind researchers warn that irresolvable public disagreement over AI consciousness could become a source of political and social conflict. - Decrypt

Google DeepMind invests $75 million in A24 to develop AI filmmaking tools for the studio's 20-person Labs team. - CNET

Google launches a 12-week AI incubator for alumni, offering up to $450,000 in credits and funding per startup - Bloomberg

Florida State University reports that Google NotebookLM improved student grades and study habits within weeks of its campus launch. - Google

Google predicts humans will stop directly using data platforms within three to five years as AI takes over queries. - theregister

Oracle shed 21,000 jobs in 12 months — 13% of its workforce — blaming AI-driven operational changes - CNBC

WEF says AI will displace 92 million jobs globally by 2030 but create 170 million new roles, a net gain of 80 million. - Fast Company

GM cut 1,000 Detroit factory workers and installed 50 AI-powered collaborative robots, drawing UAW grievances - Yahoo

As AI inference strains GPU memory, Nvidia is formalizing a new 'CMX' storage tier to bridge high-bandwidth memory and disk for key-value cache - VentureBeat

Cool cache: Nvidia's Rubin liquid cooling cuts AI data center water use to near zero by running chips at 45°C - The Verge

Meta-data breach: Meta pauses tracking employees to generate training data after the data was exposed internally. - WIRED

Spec-ulation: Meta launches new $299 AR glasses. - Meta Newsroom

REI says Meta auto-enrolled it in an AI tool that distorted a vendor image, generating a malformed bicycle ad on Instagram for a week. - Fast Company

Meta's Oversight Board calls for stronger protections against AI-generated sexualized deepfakes targeting ordinary users. - Engadget

AI PACs have poured $49M into New York's NY-12 primary, framing the race as a referendum on whether AI can be regulated - the Guardian

Workday in court: Judge ruled Workday must face the first major lawsuit targeting AI screening software's bias against Black, female, and older job applicants - TNW | Business

London's Metropolitan Police will install fixed facial recognition cameras in the West End and Soho by Christmas, following a pilot that recorded 173 arrests - the Guardian

From Russia With Slop: Russian documents reveal Project 2026, a plan to seed fake Wikipedia-style sites and think tanks to shape AI chatbot training data - Bloomberg

California drivers file class action accusing BP, Circle K, and others of using AI to fix gas prices. - the Guardian

Z.ai's GLM-5.2 open model with 744B parameters can now run locally via Unsloth Dynamic GGUF quantization at 1/5 the price of Opus. - Unsloth AI

Fast, cheap, or good, pick one or two: Head-to-head test shows Claude Opus 4.8 built a cleaner 3D browser game faster than open-weights GLM-5.2, at higher cost. - Tech Stack Ups

GLM-5.2 cost a fraction as much. Opus finished in half the time and shipped a cleaner game.

Loop dreams: Claude Code creator Boris Cherny calls agentic loops as significant a shift as the move from manual coding to AI-generated code. - TechCrunch

Anthropic engineers ship 8x more code per quarter than in 2021–2025, shifting the engineering bottleneck from coding to verification. - 24/7 Wall St.

Claude Code's extended thinking logs contain encrypted, inaccessible reasoning summaries rather than the full model reasoning that drove agent behavior. - Patrick McCanna

AI-generated code has 2.7 times the vulnerability density of human-written code, and 58% of developers deploy it without testing - Hacker Noon

Companies are reining in AI spending after years of unchecked growth, BCG projects budgets will still double to 1.7% of revenue in 2026. - Business Insider

CFOs are becoming AI budget czars, setting per-employee token limits and vetting vendors as AI spend hits $600 per engineer per month at some firms - Business Insider

Token gestures: Enterprises are deploying agentic AI faster than governance frameworks can keep pace, leaving most organizations exposed to significant operational risk. - TechRadar

As AI agents move money and push code autonomously, experts call for tamper-proof audit trails and hardware-backed identity checks - Fortune

Research frames prompt injection as LLM role-perception failure, arguing robust defenses require correct source identification rather than memorizing known attacks. - DeepFabric

Medicare's new AI prior-authorization program WISeR is generating errors, long wait times, and payment delays of six to eight weeks - CBS News

Nvidia’s Kimberly Powell: Healthcare is adopting AI three times faster than any other US sector, with tools that could free up to 40% of clinicians' time from paperwork - Financial Times

Weight, there’s more: Eli Lilly is building an AI drug discovery platform it calls an "App Store for scientists," funded by its Mounjaro profits - Financial Times

An AI-enhanced ECG tool called EchoNext detected hidden structural heart disease in a patient who went on to receive a heart transplant after further investigation - Nature

Cooley law firm launched an AI contract-review tool for Y Combinator startups to prevent risky agreements created without legal counsel - Business Insider

Cooley CEO argues Big Law must redesign its business model around AI or risk being disrupted by legal tech firms. - Fortune

UK's Garfield AI wins its first court case preparing legal documents for £400 to win £7000 judgment, cutting costs that often deter small businesses - the Guardian

Nebraska attorney James Hamilton surrenders law license after fabricating legal documents and court decisions using AI. - Above the Law

ByteDance unveils Seedance 2.5 with 4K video generation. Seedance 2.0 rollout was paused after entertainment industry uproar. - The Information

Qualcomm is nearing a ~$4 billion deal to acquire Chris Lattner’s Modular, an AI infrastructure software startup - Bloomberg

Small circuits, big machines, big bucks: ASML's new $400M extreme-ultraviolet chip-printing machine took 16 years and $10B to build; it can now print at 8nm - MIT Technology Review

Open-source AI agents OpenClaw and Hermes Agent help solo founders automate workflows with persistent memory and autonomous task execution. - TechRadar

Prosus launches ToqanClaw, a privacy-compliant OpenClaw-style AI agent platform for European merchants - Bloomberg

Microsoft Agent Framework roadmap adds file access, manual approval gates, and Azure AI Foundry integration for enterprise agents. - 4sysops

Satya Nadella argues enterprises must build 'token capital' — AI capabilities they own — or risk falling behind rivals who do - Fast Company

Microsoft's massive Texas datacenter deal locks in 20 years of gas turbine emissions as AI infrastructure demand surges. - theregister

AI use in research triples paper output and citations but is shrinking scientific diversity and collaboration, Nature reports - Nature

A for AI effort: Amateur linguist says he cracked Minoan Linear A with the help of Claude Code - aiclambake.com

Best Boy: bot, uncrededited: Hollywood studios are secretly using AI in post-production on roughly 80% of projects without crediting the technology. - Deadline

New York Times publisher A.G. Sulzberger called Google, Meta, and OpenAI thieves at a global news media conference. - The New Republic

Oxford mathematics professor John Lennox warns AI could enable authoritarian control and dangerous quasi-religious worship of technology. - The Telegraph

Thousands of data workers in Asia, Africa, and South America provide the human annotation that powers AI tools at the 2026 FIFA World Cup. - Rest of World

Robot.com is deploying R-noid, a wheeled humanoid for industrial and food service tasks, with 70% initial autonomy across a dozen customers - Business Insider

AI marks its 70th anniversary since the 1956 Dartmouth conference, with adoption rates surpassing any prior technology. - IEEE Spectrum

Follow the latest AI headlines via druce.ai on Bluesky