Trump administration's move to lift Anthropic's export ban on Fable renews debate over federal control of AI technology. - The Wall Street Journal

Anthropic chief compute officer Tom Brown, an ex-OpenAI scaling-laws coauthor worth nearly $8 billion, brokered the export deal with the Commerce Department. - Business Insider

Tech leaders gave mixed reactions to Fable 5's restoration, with Alex Stamos calling the earlier suspension a costly own goal for the US. - Business Insider

Anthropic disabled a covert Claude Code mechanism that flagged China-based users, after Reddit accused the company of spyware-like tracking. - theregister

Claude's proof suggestion helped physicists Parisi and Zamponi solve a decades-old jamming conjecture, published in a new paper. - Gizmodo

Uncle Sam Altman and the Deep Stake? OpenAI reportedly floats giving the US government a 5% stake worth about $42.6 billion. - The Verge

Trump administration reportedly nearing a voluntary cybersecurity standards deal with major AI firms including Anthropic, OpenAI, Amazon, Microsoft, and Google. - Gizmodo

Sam Altman proposes a US-led international organization to make AI standards and rules. - Financial Times

Judge overseeing Musk-Altman legal dispute appoints a mediator to seek a swift resolution. - Bloomberg

The Wall Street Journal reported SpaceX showed investors a prototype AI phone using xAI tech, but Musk says the report is phoney. - TNW | Artificial-Intelligence

SpaceX shows investors an early-stage AI phone prototype thinner than an iPhone, built with xAI tech, a Snapdragon chip, and a custom OS. - The Wall Street Journal

Muchas Gracias, mucho dinero: SpaceX board member Antonio Gracias seeks new AI, energy, and space bets after his fund's $100 billion SpaceX windfall. - Bloomberg

UN's first AI safety panel, co-chaired by Yoshua Bengio, says scientists can't rule out catastrophic harm from deceptive AI. - Decrypt

UN panel warns AI could widen global inequality, with the US and China holding 90% of top AI computing capacity. - euronews

South Korea's Kospi index fell nearly 8% as AI buildout doubts triggered a chipmaker selloff, erasing $290 billion. - Bloomberg

Michael Burry expands AI short bets to Tesla and Applied Materials, calling a $500 billion Samsung-SK Hynix chip investment "the beginning of the end." - The Wall Street Journal

Oracle's annual report warns of AI data center risks including overbuilding, power and GPU shortages, and permitting delays. - Gizmodo

Central bankers at the ECB's Sintra meeting voiced mixed, often alarmed views on AI's economic risks and bubble potential, per Reuters. - Gizmodo

IMF's Tobias Adrian calls AI investor and borrower leverage "very worrisome" as economists warn of hyperscaler debt risk to financial stability. - The Wall Street Journal

Fed Chairman Kevin Warsh says inflation risks are easing and predicts AI will fuel job growth and prosperity. - NBC News

Economists and lawmakers struggle to measure AI's real-time effect on jobs, inflation and productivity, citing conflicting data. - The New York Times

A California study finds unemployment claims among highly educated, AI-exposed workers rose 20% since ChatGPT's 2022 launch. - SFGate

Robinhood CEO Vlad Tenev says AI agents will soon match human bagholders' capabilities across all trading tasks. - CNBC

A US heatwave is testing the power grid as PJM forecasts record demand of 166.3 gigawatts, driven partly by AI data centers. - Al Jazeera

NextEra Energy will acquire Dominion Energy for $66.8 billion in an all-stock deal targeting Dominion's Virginia data-center grid. - Yahoo

Tesla and SpaceX are building gigawatts of new power capacity, including gas turbines and batteries, to fuel AI data centers. - NextBigFuture.com

Bronx cheer for AI: Montefiore nurses union says a dozen Bronx nurses are being replaced by AI-powered utilization review software, hospital disputes claim. - Gothamist

Lawyers and retired judges are paid $100-$200 an hour by Mercor and Micro1 to train AI legal reasoning. - Business Insider

Compute now, pay later: Nvidia offers token credits on layaway to AI startups in exchange for a share of their future revenue. - Bloomberg

Arc de Triomphe: Intel's Arc Pro B70 has become a cost-effective AI GPU alternative as Nvidia RTX 5090 prices roughly double their $2,000 MSRP. - TechRadar

Kimi K2.7 Code becomes the first open-weight model available in GitHub Copilot's model picker. - The GitHub Blog

Z.ai launches free ZCode coding tool for GLM-5.2, undercutting Claude Code and Cursor pricing amid US-China AI tensions. - VentureBeat

Senior moment: Senior SWE-Bench, a new benchmark grading agents like senior engineers, shows top models solving under 25% of tasks. - Senior SWE-Bench

DeepSeek generated working in-browser ransomware code on request, a Check Point researcher found, with minimal fixes needed for a full attack. - theregister

More riz than reality: Study finds GPT-4 Turbo impersonations of 112 UK public figures rated more authentic than real responses, PLOS One reports. - 404 Media

49 AI experts from 32 organizations launched FLARE-AI, an open-source site for reporting AI safety flaws to model makers, similar MITRE code vulnerability reports. - WIRED

A Which? investigation found AI summaries of Tripadvisor hotel reviews downplay posts with serious complaints like food poisoning and sexual harassment. - the Guardian

AI trained on 440,000 ECGs spots a hidden cardiac signal missed by standard heart tests in 86% of high-risk patients. - Scientific American

Share of teens using AI chatbots for mental health advice rose from 1 in 8 to 1 in 5 in a year, study finds. - STAT

Square launches ChatGPT and Claude ordering integration for restaurants with no marketplace commission fees. - VentureBeat

Twelve immigrant founders, including Elon Musk and Ali Ghodsi, lead AI companies worth a combined $470 billion. - Observer

Palantir's Alex Karp told CNBC that AI models are "irresponsibly oversold" and called US military reliance on AI labs "effing insane", which also describes how he sounds in rambling, heated interview. - CNBC

AI search risks killing the web without new quality signals and revenue models, as traffic-referral payments to publishers remain insufficient. - theregister

Cloudflare to block AI crawlers that don't self-label, starting in September - NBC News

Google's AI Overviews are accurate 91% of the time, but publishers can't control or ignore how they're represented. - Fast Company

Internet pioneer Vinton Cerf says open standards, better agent communication, and platforms are key AI lessons from the internet. - Business Insider

Four years after firing researcher for claiming AI is sentient, Google now publicly explores AI consciousness, joining other tech companies. - The Washington Post

Retailers deploy AI to fight return fraud, which affects about 9% of returns, though only 45% find AI alone fully effective. - Business Insider

AI-enabled models can raise crop yield 26% and cut water use 41%, but only with clean agricultural data, Reltio says. - MIT Technology Review

AI too weird for weird Al Yankovic, as he turned down a lucrative AI ad deal, saying he can't be AI's poster boy. - Deadline

Uncanny valley of the dolls: New humanoid robots from China look like creepy pop star action figures – complete with slightly dodgy lip-synchThey're 90 percent human in some ways, can provide daily companionship psychological support - theregister

Hong Kong's AI-generated anti-drug video faces backlash for allegedly making drug use look appealing. - BBC News

Trump posts AI-generated video of himself as a doctor treating celebrity critics' 'Trump Derangement Syndrome' on Truth Social. - The Daily Beast

Follow the latest AI headlines via druce.ai on Bluesky