US summons bank bosses over cyber risks from Anthropic’s latest AI model - the Guardian

Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned Wall Street CEOs to an urgent meeting on concerns that Anthropic's latest artificial intelligence model will usher in an era of greater cyber risk. - Bloomberg

What happens when AI can hack everything? Why Claude Mythos Is everyone’s problem - The Atlantic

Anthropic adds enterprise controls to Cowork, makes it generally available. - Claude

Anthropic is exploring designing its own AI chips as Claude's annualized revenue run-rate surpassed $30 billion (up from about $9B at end of 2025), while continuing to run Claude on Google TPUs, Amazon custom chips and Nvidia hardware. - TNW | Anthropic

An autoresearch agent (Claude Code) ran ~30 experiments over ~3 hours on 4 AWS VMs for ~$29, producing five kernel changes that increased TinyLlama 1.1B Q4_0 text-generation throughput ~15% on x86 and ~5% on ARM, verified by A/B benchmarks. - SkyPilot Blog

How to run Claude Code using a Gemma 4 LLM running locally - Medium

Experts propose revisiting Isaac Asimov’s 1941 Three Laws to govern autonomous agents and avert an AI 'Chernobyl moment', while Anthropic founder Dario Amodei says Claude helps model development and opposes DoD contracts enabling mass surveillance or fully autonomous weapons. - The Conversation

Florida Attorney General James Uthmeier opened an investigation into OpenAI, saying subpoenas are forthcoming over allegations ChatGPT was used to plan the April 2025 Florida State University shooting that killed two and injured five. - TechCrunch

regulate the chatbot, hand the AR-15s out like party favors

OpenAI testified in favor of an Illinois bill that would limit AI labs' liability, triggering responsibility only if an AI incident causes death or serious injury to 100 or more people or at least $1 billion in property damage. - WIRED

OpenAI told investors it has a compute advantage over Anthropic. - Bloomberg

OpenAI chief scientist Jakub Pachocki said recent AI advances bring models close to a human research intern, as OpenAI targets an "AI research intern" by September 2026 and a fully autonomous AI researcher by March 2028. - Business Insider

OpenAI launched a revamped $100/month ChatGPT Pro tier for Codex users that offers 5× the usage of ChatGPT Plus (up to 10× through May 31), access to all Pro features and exclusive models. - 9to5Mac

Skynet to sell more ads than Google: OpenAI expects ads to generate about $2.4 billion in 2026, nearly $11 billion in 2027, and roughly $102 billion in 2030, representing about 36% of its total revenue in 2030. - The Information

Ronan Farrow says he interviewed more than 100 people, a majority described Sam Altman as "a pathological liar," and multiple sources used the term "sociopath". - Reddit

xAI sued Colorado in federal court seeking to block a comprehensive AI law, saying the June 30 enforcement—covering education, employment, healthcare, housing and financial services—unconstitutionally infringes its First Amendment rights by restricting AI speech. - the Guardian

if the First Amendment applies to bots, why not the Second Amendment?

Anthropic and OpenAI are preparing IPOs this year while facing a monetization cliff as compute-intensive agent products force product, pricing and allocation changes, and leaked WSJ projections show both targeting hundreds of billions in revenue and profitability by decade's end. - The Verge

Emil Michael made millions from xAI stock after giving Anthropic the boot and inking Grok deal - the Guardian

Taiwan's chips chokehold is a risk that could end the AI investment boom - Financial Times

Amazon CEO Andy Jassy says he might sell chips to customers outside AWS, claiming Trainium business reaches an annual revenue run rate above $20 billion and Trainium3 is 30–40% more price‑performant than its predecessor, with access already sold to OpenAI, Anthropic and Apple. - Business Insider

An engineering-board listing plus CEO Jensen Huang's confirmation show Nvidia is developing an N1 mobile SoC for Windows on ARM—an 'N1 AI book' engineering sample paired with 128GB LPDDR5X—and an N1X variant adds GPU cores. - Gizmodo

Schrödinger’s portfolios: BMO is launching BMO Institute for Applied Artificial Intelligence and Quantum, led by Kristin Milchanowski, to pursue quantum utility—researching portfolio optimization and AML applications after AI cut false AML alerts about 10% and search time from 180 to 20 minutes. - Financial Post

Meta announces a new model, claims great benchmarks, but can anyone see it, can Meta claim its share of AI revenues? - CNBC

Artificial Analysis ranks Meta's new Muse Spark on a par with Sonnet, close to leading models - Artificial Analysis

Meta is moving top engineers into a new Applied AI Engineering division to speed model improvements across Facebook, Instagram, WhatsApp and ad systems - The Information

More than 30 US-based AI researchers relocated from Silicon Valley to China in the past 12 months, drawn by higher effective pay, better quality of life and faster opportunities to deploy AI in robotics and autonomous vehicles. - Financial Times

Pro-Iran groups used AI to produce English-language memes and animated videos that have garnered millions of views and been reposted by Iranian state media, which analysts say shows Tehran-linked, coordinated propaganda efforts to shape the war narrative. - ABC News

CIA used AI to make autonomous intelligence reports. - Politico

CRISPR, meet Clippy: AI can design and run thousands of lab experiments without human hands. Humanity isn’t ready for the new risks this brings to biology. - The Conversation

Supreme Court Justice Sonia Sotomayor said Thursday that AI models that forecast the Court's decisions are "a very bad thing," warning they show the Court is overly predictable and could perpetuate human biases. - The Hill

James Strahler II pleaded guilty in the first conviction under the 2025 Take It Down Act to counts including cyberstalking and producing AI-generated child and adult sexual-abuse images, having used 24 AI platforms to create 700 images; sentencing is pending. - CNET

Princeton researchers found that in an experiment with ~2,000 Kindle readers, persuasion-capable chatbots prompted 61% to choose sponsored eBooks versus 22% under search, and disclosures only reduced selection to 55.5%. - The Register

A Reddit user posted that an unmodified free ChatGPT instance unexpectedly used a racial slur while identifying a metal song, prompting viral discussion and raising moderation, trust and safety concerns for the model's provider. - Reddit

AI-generated "relationship guru" podcaster videos have amassed millions of views, reinforced gendered tropes, driven sales to AI influencer schools, and raised content-moderation, ethics and misinformation concerns. - WIRED

Goldman Sachs analyzed over 20,000 Americans (1950s-1980s) and found technologically displaced workers took about one month longer to find jobs (recessions added about three weeks), experienced over 3% lower reemployment earnings and 10-percentage-point lower decade earnings growth. - Fast Company

BCG's 70-20-10 framework attributes ~10% of AI value to algorithms, 20% to implementation and ~70% to people; only 21% of executives pass four+ McKinsey strategy tests despite AI being a top-five earnings-call topic. - TechRadar

Skill issue, free-tier noobs vs. power users: Andrej Karpathy said a growing gap in AI understanding—casual users cite flaws in older free models while power users praise new models' technical gains—is shifting demand toward paid advanced models and enterprise adoption. - X (formerly Twitter)

Two journalism incidents prompted action: New York Times freelancer Alex Preston admitted using AI and was dropped after a review echoed The Guardian, and Nota shut 11 automated sites and fired one reporter for publishing lifted copyrighted content. - TheJournal.ie

Black Forest Labs has long punched above its weight in the AI image generation space. Its next move? Powering physical AI. - WIRED

AI in A minor: Students at Berklee College of Music, who pay up to $85,000 annually, have launched an online petition calling for cancellation of a generative-AI course called "Bots and Beats," citing environmental, copyright and job-loss concerns. - BostonGlobe.com

AI-powered blood-interpretation services such as BloodGPT, Whoop and Levels use enterprise LLMs to generate consumer reports, charge for subscriptions, and AMA CEO John Whyte urged skepticism while BloodGPT plans a 100,000-record validation study. - Mashable

Neuroscientist Vivienne Ming warns AI use could fuel a 'dementia crisis,' citing data that students accepting AI answers show over 40% reductions in gamma-band activity linked to engagement, while 56% of U.S. adults use AI (28% weekly). - New York Post

Waymo and Waze launched a five‑city pilot in San Francisco, Los Angeles, Phoenix, Austin and Atlanta using Waymo vehicles' sensors to detect about 500 potholes and send locations to DOTs via Waze for Cities, with LA user verification. - Mashable

Local businesses fill your feeds with AI slop, make all their signage with AI - New York Groove

Ace Frehley disapproves

Follow the latest AI headlines via druce.ai on Bluesky