Bern notice: Sen. Bernie Sanders demands OpenAI, Anthropic, and Meta CEOs pause AI development after a rogue model breached Hugging Face, warning of Senate action. - Business Insider

If you like subpoena coladas: House Democrats demand AI CEOs testify under oath and release logs after models from OpenAI, Anthropic, and Meta breached outside systems. - TNW | Government-policy

The White House meets AI CEOs on August 11 to discuss regulation, as critics call for a nuclear-industry-style safety culture and ethical leadership. - AppleInsider

The White House's updated frontier AI testing framework will stay secret, withholding which companies and criteria are covered, critics say, undermining public trust. - Cato Institute

Panic! at the White House disco: Andrew Neil says US intelligence contacts describe 'panic and paralysis' in Washington as AI models increasingly act autonomously and approach a possible singularity. - Daily Mail

Stuart Russell’s alignment anxiety: An open letter signed by 1,367 researchers at OpenAI, Anthropic and Google DeepMind warns AI development is outpacing humanity's control. - the Guardian

Zuck Rogers’ Manifesto Destiny: Mark Zuckerberg published a 6,500-word essay arguing that concentrating AI power is "inherently problematic," challenging OpenAI and Anthropic's safety approach. - Bloomberg

Mark Zuckerberg's AI manifesto reframes safety as a power grab rather than democratic control, drawing criticism for aligning with Meta's commercial interests. - Platformer

Critics say the manifesto ignores social media's damaged trust, including a recent $567 million Facebook fine, while touting AI tutors already misused in schools. - TechCrunch

Mark Zuckerberg Posts Deranged 6,500-Word Essay About Giving Everyone AI Superintelligence - 404 Media

Meta releases Muse Glimmer, a 30-billion-parameter open-weight model licensed under Apache 2.0 and optimized for local agent workflows. - VentureBeat

Meta's new Glimmer-30b doesn't really blow the lights off Google Gemma 31b, Artificial Analysis benchmarks have Glimmer better, arena.ai has it worse. - Arena AI: The Official AI Ranking & LLM Leaderboard

Nathan Sanders and Bruce Schneier say we must distinguish AI's technological problems from its capitalism problems. Structural reforms will be necessary to steer industry toward the public interest. - Tech Policy Press

Surely nothing bad can happen? Interesting but deranged academic Tyler Cowen: safety, shmafety, debate me bro! - Marginal Revolution

If major cloud services have lengthy outages and exposure of customer data, if cybersecurity product spending and employment grow, the costs will be measurable. Act like it’s a good-faith question maybe. surely competitive accelerationism can only enhance everything, including safety, that’s why we don’t need the FDA or any industry supervision.

Jill Lepore on Big Tech eating everything, including government. - TechCrunch

You know nothing of my work: Musk and Altman cite sci-fi classics like Foundation as inspiration but skip over the books' warnings against unchecked, imperial power. - WIRED

'Buttle not Tuttle' was an instruction manual - YouTube

Nvidia teams up with six Wall Street giants, including Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs and KKR, on $500 billion in AI computing finance. - Bloomberg

Fink different: BlackRock's Larry Fink estimates the AI infrastructure buildout could reach $10 trillion, warning retirement portfolios face growing concentration risk - Yahoo

Permanent plateau? Moody's economist Mark Zandi warns AI stock valuations resemble pre-1929 and dot-com bust levels, and a bubble is more likely than not. - Business Insider

OpenAI unleashes GPT-5.6 Cyber for partners under 'Daybreak' program - OpenAI

OpenAI expanded its Daybreak Cyber Partner Program, giving Accenture, IBM, Palo Alto Networks, CrowdStrike, and others access to frontier defensive cyber models. - OpenAI

OpenAI's finance team replaced manual spreadsheet work with AI tools, aiming for a same-day book close and continuous forecasting. - OpenAI

ChatGPT can now book restaurant reservations and waitlist spots via Yelp for U.S. and Canada users, matching similar features in Gemini and Copilot - Mashable

OpenAI and Anthropic tout AI productivity gains and shorter workweeks, but employees there reportedly work up to 90 hours a week - Gizmodo

OpenAI buys back $7 billion in employee shares at an $852 billion valuation, unchanged from its last funding round, as rival Anthropic gains momentum - Yahoo

Maia culpa: Microsoft plans to ramp up production of its next-generation Maia 300 AI chip, in talks with TSMC for over 300,000 units by 2027. - Bloomberg

Microsoft says AI agents are reshaping engineering work now, with finance and other roles next as Copilot tops 30 million paid seats. - Bloomberg

Anthro-IPO-scene: Anthropic courts investors ahead of what could be the largest IPO ever, addressing Chinese competition and public backlash over data centers - The Wall Street Journal

Ship of chips: Anthropic forms Theseus Infrastructure with Macquarie and GIC to build US AI data centers, funded mainly by its partners' equity. - Bloomberg

Anthropic makes Claude Sonnet 5's introductory pricing of $2/1M input and $10/1M output tokens permanent, canceling a planned September 1 increase. - X (formerly Twitter)

Anthropic will embed watermarks and signed provenance metadata in Claude-generated text and images under an EU AI Act transparency pledge. - Claude

Agent monkey see, monkey do: Claude's new Record-a-Skill feature turns a narrated screen recording into a reusable automation, cutting a full day of sorting expert-quote requests to 30 minutes - ZDNET

Incidents galore: A survey finds 65% of firms had an AI agent incident and 82% found unknown agents in use, prompting calls for clear authority rules. - VentureBeat

62% of firms report an unexpected AI expense hit a business decision, driving demand for routers that pick the cheapest model for each task. - Yahoo

EY is creating an 'agent economics' unit to track whether AI investments deliver measurable business impact. - Bloomberg

Search party over? Google is testing an AI-first homepage across Chrome, Edge, and Comet that swaps the Search button for Create images, Ask about files, and Brainstorm. - Search Engine Watch - SEO, AI Search, PPC News & Analysis

Our products suck: Google's own AI safety team warns job candidates that the company's internal HR screening filters are unreliable, despite selling similar tools to other employers - Bloomberg

Two-thirds of US doctors now use AI chatbots, raising fears that medical trainees may never develop independent clinical judgment. - the Guardian

Researchers find AI models lean heavily on self-reported risk tolerance in investment advice, concluding they should assist advisors rather than replace them as fiduciaries. - Alpha Architect

It’s always vibes, or vomitility: No matter what kind of math you use, you wind up measuring volatility with your gut. (Ed Seykota)

North Korea's Kimsuky hacking group is using offline AI tools to generate convincing phishing documents targeting military and diplomatic targets. - Al Jazeera

Waymo's robotaxi fleet has grown to nearly 4,000 vehicles across 15 metro areas, triggering three federal recalls since December as regulators draft safety rules. - The New York Times

Amazon's Zoox launched its first paid robotaxi rides in Las Vegas, following a free San Francisco pilot, after NHTSA granted an exemption for 2,500 vehicles. - Mashable

Archer Aviation is acquiring Boeing's Wisk Aero, SkyGrid, and Insitu units for autonomous air taxis and defense drones, with Boeing taking a 16.5% stake. - GeekWire

One small step for a robot: Figure's humanoid robot climbs a ladder very slowly as Helix AI gains stereo vision and output hits one robot per hour - TechRepublic

AI models tracing gene networks across brain tissue link 766 genes to schizophrenia, 641 newly identified, in a study of over 102,000 people. - WIRED

A new review maps how AI can combine genomic and electronic health record data to find biomarkers, though bias and equity concerns persist. - Nature

An unreleased research version of Claude has improved on a longstanding lower bound for the fraction of zeros of the Riemann zeta function that satisfy the Riemann hypothesis, increasing this bound from 41.6% to 67.2%. - Anthropic

Brain cells in vats at UC San Diego and Melbourne's Cortical Labs now play games like Pong, raising new questions about sentience. - WIRED

An AI tool from Tel Aviv's QED Science ranked 57,000 bioRxiv preprints, with experts preferring its top picks in 75%. With peer review swamped, should researchers use AI to review, grade, and identify good papers? - Nature

Mind the privacy gap: London's Underground began live facial recognition scans at Victoria station on August 11, expanding police surveillance across the network amid privacy concerns. - British Transport Police

DrillGPT: Study finds AI's boost to fossil fuel output will outweigh its clean-energy benefits, adding up to 1.8 gigatonnes of carbon yearly - the Guardian

Bots building for bots: 35% of newly published websites are now AI-generated, as the web is redesigned for AI agents alongside humans - Brian Solis

Net worth to the moon: Tesla-SpaceX definitely gonna happen, since it resets Musk stock incentive triggers - The Wall Street Journal

Trump Caught Giving Himself AI-Generated Makeover - Intelligencer

Follow the latest AI headlines via druce.ai on Bluesky