Moonshot heard round the world: Kimi K3 triggered a global selloff, pushing US semiconductors into a bear market and pulling Nasdaq down 1.4%, S&P 500 down 1%. - The Wall Street Journal

Moonshot's 2.8-trillion-parameter Kimi K3 model beat Claude Fable 5 and GPT-5.6 Sol on coding benchmarks, narrowing the US-China AI gap to about six months - The New York Times

Kimi K3 is now the largest open-weight AI model at 2.8 trillion parameters, beating Claude Fable 5 on some benchmarks while priced far below rivals. - Business Insider

Tech and policy figures split on Kimi K3, with David Sacks calling it concerning and Aaron Levie and Gavin Baker calling it a win for competition. - Business Insider

GLM-5.2 is the most cyber-capable open-weight model, closing the gap with Anthropic's Opus to 4 to 7 months, versus 6 to 10 in 2025 - AI Security Institute

The 2026 World AI Conference in Shanghai issued a chair's statement outlining 13 principles for global AI governance and cooperation. - GlobalSecurity.org

China's foreign ministry dismissed US accusations, including from Anthropic, that Chinese AI firms illicitly distill foreign models, calling the claims misguided and counterproductive - Bloomberg

The White House is asserting new control over which companies can access frontier AI models from Anthropic and OpenAI. - CNBC

Google's AI capital spending could rise further as capacity constraints leave a $462 billion Cloud backlog unmet, Sundar Pichai says. - Yahoo

Google's Gemini 3.5 Pro model, promised for June, remains unreleased months later after missing internal coding-performance targets, rattling investors and eroding Alphabet's stock - 9to5Google

Google shifted Gemini's pricing model to bill by computational cost instead of fixed request counts, complicating usage limits across paid tiers - WIRED

Google shares 13 hands-on tutorials for building, scaling, governing, and optimizing agents on its Gemini Enterprise Agent Platform. - Google Cloud Blog

OpenAI's new ChatGPT Work mode delegates multi-step tasks end to end, with most small-business leaders now confident handing high-stakes work to AI agents - Entrepreneur

OpenAI will notify parents when a teen's ChatGPT account is banned for violent content, following criticism tied to a 2025 mass shooting - Engadget

OpenAI updated its ChatGPT Mac app to restore easier access to chats after user complaints about a confusing redesign. - 9to5Mac

Apple's AI-enhanced Siri takes a utility-first approach, organizing emails and notes rather than trying to chat like ChatGPT or Claude - PCWorld

Apple's new Siri AI gave direct answers to real-world questions on Apple Watch, where the old Siri often just found web results. - CNET

In reversal, Anthropic will not remove Claude Fable 5 from Max and Team Premium. Half usage limits from July 20; Pro users shift to metered API pricing. - X (formerly Twitter)

Anthropic researchers say Claude shows a brain-like "global workspace" pattern (a hub theory of consciousness), but a philosopher says key differences remain - SingularityHub

Anthropic wanna get Meta-physical: Anthropic and Meta are discussing a $10 billion data center leasing deal, echoing Anthropic's existing SpaceX capacity agreement - SiliconANGLE

Anthropic is in early talks to lease up to $10 billion of computing power from Meta over two years. - CNBC

Friends with compute benefits: SpaceX is in talks to lease its data center capacity, some running xAI models, to the Pentagon for billions - Futurism

SpaceX is in talks with the Defense Department to sell billions of dollars in AI data-center capacity for the Pentagon's AI push. - The Wall Street Journal

Aerial imagery analysis of SpaceX's Colossus 2 site shows 440,000 B200 and B300 AI chips active as of early July 2026. - NextBigFuture.com

The NAACP sued over air-permit-free gas turbines powering xAI's Mississippi data center, citing toxic pollution risks to residents - The Daily Beast

The IMF war-gamed how AI could shrink income tax revenue with experts from Google DeepMind, Rand, and the Federal Reserve under a secrecy rule - Bloomberg

The Fed's Beige Book finds AI data center construction straining supplies of electricians, memory chips, and steel amid a broader shift to capital-driven growth - Yahoo

Microsoft is replacing OpenAI and Anthropic models with in-house MAI models in Outlook and Excel, citing Anthropic's rising costs, to cut reliance on rivals - Yahoo

JPMorgan CEO Jamie Dimon warns Anthropic's Claude Mythos model poses risks comparable to giving ballistic missiles to individuals. - TechRadar

Going big in Japan: Nvidia and Japan's Noetra consortium are building a 140-megawatt national AI factory with 27,500 Vera Rubin GPUs, backed by up to $6.1 billion. - TechSpot

AI now accounts for over 80% of newly leased San Francisco office space, pushing loan delinquencies to half the national rate. - CRE Daily

Hot tub compute machine: Architects are redesigning data centers with public bathhouses and parks as 71% of Americans oppose new sites nearby - Business Insider

An autonomous AI agent ran a multi-stage breach at Hugging Face, accessing internal datasets and credentials before the company patched flaws and rotated tokens - 4sysops

Amazon Web Services launched Loom, an open-source platform for building and governing AI agents with security controls. - Yahoo

AWS detailed an agentic classification architecture for regulated tasks spanning law enforcement, healthcare, and financial services. - Amazon Web Services

Amazon's new AI staffing enforcement system prompted managers to override or shut down its recommendations during early pilot testing. - Yahoo

OWASP framework urges least-privilege access and sandboxing to stop AI agents from prompt injection and tool misuse attacks. - MachineLearningMastery.com

Airline thesis: Steve Eisman warns AI giants' trillion-dollar spending lacks pricing power, comparing the sector to capital-intensive airlines with no moat. - Yahoo

Investor Chamath Palihapitiya predicts cheaper AI models from Meta and xAI will force a price collapse in premium offerings like OpenAI and Anthropic. - Yahoo

Index Ventures co-founder Neil Rimer predicts voluntary or forced redistribution of AI-driven wealth as billionaire philanthropy declines. - TechCrunch

OECD ranks London the world's most AI-exposed major city, with UK forecasters warning 3.4 million jobs could vanish within a decade - The Independent

Indeed's chief economist says the real US labor problem through 2032 is nearly 6 million retiring Baby Boomers, not AI automation - Fortune

Gig economy, god tier: Startups like Mercor and Surge AI pay contractors up to $1,000 an hour to grade AI outputs, but work dries up fast. - Yahoo

Customertimes cofounder Max Votek says AI savings are often absorbed by token and infrastructure costs, not diverted to executive bonuses as commonly assumed - Business Insider

Physical AI startups building self-learning factory and warehouse robots raised over $8 billion in H1 2026, more than double all of 2025 - Yahoo

Rock’em, sock’em: Humanoid MMA Fight Ends in Decapitation - Common Dreams

China's BrainCo demoed an EEG headset it calls the first integrated brain-to-robot platform, controlling robotic arms, dogs and humanoids by thought alone - South China Morning Post

Ex-missile-engineer Alex Toussaint's startup Tornyol built a $1,100, 40-gram AI drone that uses sonar to detect and intercept mosquitoes mid-flight. - TechRadar

Nobel laureate Jennifer Doudna's lab used AI to design a new gene-editing enzyme that worked in human, plant, and bacterial cells - FierceBiotech

Google, Microsoft and IBM are advancing quantum computing as AI's likely successor, with Microsoft targeting a scalable quantum computer by 2029. - Metro

Tech executives are hiring armed bodyguards as anti-AI backlash turns violent, including a foiled attack on Sam Altman's home - Yahoo

Esther Perel warns AI is deepening workplace social atrophy by replacing human conversations with chatbot interactions. - Fortune

The rent is too damn hallucinated: NYC mayor Mamdani orders landlords to disclose AI-altered rental listing images in new Rental Ripoff Report. - PetaPixel

Honey, it’s just a chatbot: Prenup attorneys report growing demand for clauses defining AI chatbot relationships as infidelity, though enforcement remains legally untested - Business Insider

Spotify removed 75 million AI-generated "slop" tracks in 2025 as bots upload roughly 100,000 songs a day. - TechSpot

TikTok is testing an opt-in tool that scans for unauthorized AI-generated content using a creator's likeness. - The Verge

Visa's new AI assistant compares a cardholder's spending against 300 billion annual transactions to flag savings - The Street

EY developed a multimodal RAG framework that links text, charts, and images through a knowledge graph to improve enterprise AI answer accuracy. - SiliconANGLE

Face AI upgrades its video face swap tool with faster processing, better tracking, and sub-minute rendering for social creators. - TNW | Artificial-Intelligence

Todd Blanche draws rebuke from Trump-appointed judge after using AI to lie - Raw Story

Reconstructing a video of a famous lost Pélé goal that was never recorded, using AI from radio and eyewitness accounts, for the Pélé museum. Starts at 7:15, it's kind of entertaining. - YouTube

Follow the latest AI headlines via druce.ai on Bluesky