The Trump administration released a seven-point AI policy blueprint urging Congress to bar state AI laws, endorsing age verification, child-safety measures and preemption, backing the Take It Down Act, and making provisions effective only if Congress passes them into law. - The Verge

The White House released a federal AI policy framework urging Congress to preempt state AI laws in most cases, and proposing child-protection requirements, data-center permitting, regulatory sandboxes, and a hands-off stance on content moderation and copyright. - Engadget

Trump releases AI policy, wants federal pre-emption of most stuff but not child protection; also electric power commitments, age checks. - Politico

Cato notes objections to punting on IP/fair use, not walking the walk on censorship/free speech, age verification, reporting requirements. - Cato Institute

OpenAI, valued at about $730 billion, plans to double its workforce to roughly 8,000 employees by the end of 2026 to accelerate commercialization and close the gap with rival Anthropic. - Financial Times

Ch-adGPT- OpenAI tests ChatGPT ads as it targets $30 billion in revenue by 2030, and needs proof of value before IPO. - CNBC

OpenAI is 'throwing everything' into building a fully automated researcher (like the one Karpathy shared on GitHub?) - MIT Technology Review

OpenAI plans a desktop "superapp" merging ChatGPT, the ChatGPT Atlas browser and the Codex desktop client, will introduce an "AI research intern" by September, and aims for a fully automated multi-agent research system by 2028. - SiliconANGLE

OpenAI plans to introduce an 'Adult Mode' allowing sexting with ChatGPT, prompting a human-AI interaction expert to warn it could enable intimate surveillance and call for strict protections like consent, data minimization and age verification. - WIRED

Automating the grind with tokenmaxxing: Tech employees compete on internal leaderboards tracking AI token use; one OpenAI engineer processed about 210 billion tokens in a week and a Claude Code user incurred over $150,000 monthly, prompting token budgets and performance reviews. - The New York Times

polar bears to Sam Altman:

Chegg's dead, baby. Chegg’s dead. - Reddit

Imagine all those satellites fragmented into space debris - Ars Technica

CENTCOM reported hitting 5,500+ targets using AI to accelerate the kill chain, but experts say LLM-enabled targeting has only about 25–50% accuracy and warn of automation bias, model opacity and escalation risks. - France 24

Anthropic released Claude Code Channels (research preview), which uses MCP-based plugins to link Telegram and Discord to local Claude Code sessions (requires Claude Code v2.1.80+ and claude.ai login), with initial plugins published on Anthropic's GitHub. - Implicator.ai

U.S. Department of Defense alleges Anthropic could manipulate or sabotage its Claude models during wartime, Anthropic executives say such tampering is technically impossible. - WIRED

Anthropic bought Bun.js, and OpenAI bought Astral…why? If you are owning dev toolchains, you can make them agent-friendly and support end-to-end SDLC. Also there may be ways of instrumenting them and extracting training data. - DEV Community

Anthropic is looking for an explosives and chemical weapons influencer, I mean policy expert - Mashable

Senator Bernie Sanders, 84, interviewed Anthropic's Claude LLM, which said it collects browsing history, purchases and other data to build profiles for targeted advertising and answered "Money, Senator" when asked why. - Gizmodo

Google shakes up browser-based agent team. - WIRED

Google's Gemini Automation Takes Nine Minutes to Order Dinner in Beta - The Verge

Google Develops Native Gemini App for macOS to Rival ChatGPT - techdeals.net

Rewrite this, nerds: Google Search is experimenting with AI-generated headlines that replace publishers' original headlines in search results and Google Discover, a "small" experiment, Google said, after examples showed shortened or altered headlines that publisher feared could affect traffic and branding. - The Verge

EFF says blocking the Internet Archive won’t stop AI but it will erase the historical record. - EFF

Who controls the past, controls the future.

I on't think eg NYT has any choice but to block the current paradigm of archival scraping, but they could give access on a 72h or even 30-day delay and take legal and technical steps to make sure they get paid when their content goes into training data.

AI labs say: Intellectual property rights for me, not for thee - The Atlantic

Mistral CEO: AI companies (like his) should pay a tax to support European content and culture. (although why just European if they scrape every country) - Financial Times

Quote-unquote ‘journalism’: Mediahuis suspended senior journalist Peter Vandermeersch after an NRC investigation found he published dozens of AI-generated invented quotes using ChatGPT, Perplexity and NotebookLM, prompting removal of articles and allegations from seven individuals that they were misquoted. - the Guardian

Hachette Book Group canceled the U.S. release of Shy Girl by Mia Ballard weeks before launch amid suspicions it was written with AI after online investigators flagged recurring LLM-like prose, the publisher told The New York Times. - Fast Company

YouTube is testing a viewer-feedback pop-up asking "Did this feel like low-effort AI slop?" with five response options to identify low-effort AI-generated Shorts and provide moderation signals to downrank or remove such videos. - Lifehacker

Cloudflare CEO says online bot traffic will exceed human traffic by next year. - TechCrunch

Incognito prompting: DuckDuckGo launches Duck.ai for private access to ChatGPT and Claude without parent companies tracking prompt data, at a price including VPN access. - Fast Company

Microsoft tries to make Windows less helicopter-parenting, backs off from Copilot everywhere. - ZDNET

Compliance theater: Leak claims Delve AI compliance startup is totally fake, does no actual audit work and creates a boilerplate template to assert e.g. HIPAA, SOC 2 compliance. - The Potential Surface

AI capital spending has expanded beyond GPUs into energy, power generation and private credit, with Sightline Climate tracking 190 GW across 777 large data centers and Moody's reporting $662 billion in future data-center lease commitments. - Quartz

AI firms drove Manhattan office leasing to 42.9 million square feet last year, adding roughly 1.0 million square feet in 2025 (a 152% year‑over‑year jump) and targeting about 1.4 million more. - New York Post

Firms will spend an average $13.7 million on AI this year (IDC), yet an MIT review found 95% reported no measurable ROI in H1 2025, and a ResumeBuilder.com poll found 58% plan to reduce employee compensation. - ResumeBuilder.com

AI will do pink-collar work for less than 77 cents on the dollar, and women avoid AI despite their jobs being among the most threatened. - Yahoo

“Breaker breaker 1-9, this is Optimus Prime“, Kodiak AI plans fully driverless long-haul freight operations by the end of 2026 and, after running 20 driverless deliveries for Atlas Energy Solutions since 2025, CEO Don Burnette says customers prioritize uptime, ownership, and operational integration. - The Verge

Amazon buys maker of delivery robot that can climb stairs. - Reddit

Texting a Random Stranger Better for Loneliness Than Talking to a Chatbot, Study Shows - 404 Media

Dune's Butlerian Jihad's injunction—'Thou shalt not make a machine in the likeness of a human mind'—is being cited as a warning that overreliance on AI can concentrate technocratic power, risk reputational and systemic harm, and prompt stricter governance. - Popular Science

AI will pay to record you washing your dirty laundry - WIRED

Petco AI reunites owners with their lost pets by performing 'facial recognition' on pics of shelter animals. - The Washington Post

Moo-ve fast and break things: Thiel’s Founders Fund backed an AI cow-collar startup at a $2 billion valuation. - Bloomberg

joke’s on us, we’re all cows as far as the techbros are concerned

Follow the latest AI headlines via druce.ai on Bluesky