Incident response, Oval Office tier: Altman will brief the White House this week on OpenAI's most powerful model, following its recent security incident. - TNW | Artificial-Intelligence

Sacks, lies, and export controls: Trump's AI advisers are split on strategy: Lutnick weighs incentives and export controls, Sacks wants a hands-off approach, and Bessent threatens sanctions over distillation. - WIRED

Zvi Mowshowitz: We now have more details of what happened. Every time we learn more details, it somehow makes things seem worse. - The Zvi

Hugging Face chief urges OpenAI transparency over rogue agent, support for cyber defense. - the Guardian

If OpenAI can’t stop their agents from running amok, what chance have the rest of us got?

Sam Altman says his biggest fear is a single company controlling AI — a choice, he says, between "AI authoritarianism or liberty." - Business Insider

OpenAI CEO Sam Altman says we are already "in the singularity," with AI advancing unpredictably beyond human intelligence. - Business Insider

OpenAI CEO Sam Altman says AI hasn't shortened the workweek because competitive people keep raising their own expectations. - Business Insider

Nvidia is pursuing AI infrastructure deals worth more than $750 billion, reviving fears of circular financing across the industry. - Bloomberg

Nvidia is in talks to guarantee up to $250 billion in financing for OpenAI's planned $500 billion, 10-gigawatt Ohio data center. - The Wall Street Journal

Cost to insure Nvidia's debt against default jumped as much as 0.14 percentage point Monday amid reports of $750 billion in AI deal talks - Bloomberg

Nvidia makes a "substantial" investment in Ilya Sutskever's Safe Superintelligence, boosting its compute with Nvidia GPUs and access to the next-gen Vera Rubin platform - The Wall Street Journal

Nvidia is expanding its South Korea AI push: a $500 billion letters-of-intent deal with SK Group and a $1 billion investment in Naver's data center. - The Wall Street Journal

Nvidia launched the Open Secure AI Alliance with 27 members, including Microsoft and SpaceX, but not OpenAI, Anthropic, Google, Meta, to contribute open models and cryptographic verification standards for cyber defense. - Engadget

Kernel sanders: Accomplish AI researchers found a Linux kernel flaw letting Claude Cowork's sandbox gain root access to a Mac's file system, exposing SSH keys and credentials - AppleInsider

Anthropic is the only major AI lab not backing an open-weight AI letter, drawing criticism from David Sacks, Bill Gurley, and Kai-Fu Lee - Business Insider

Anthropic on prompting new models - Claude

There are some complaints about Claude Code going off the rails, looping and not finishing simple tasks - Reddit

Publicly shared Claude conversations and Artifacts turned up again in Google search results, exposing API keys, credentials and personal addresses - 404 Media

The Eisman cometh: ‘Big Short' investor Steve Eisman dumped his Alphabet stake, warning AI hype risks a broad market correction - CNBC

Doug Kass mocks him …. Funny how people are super bullish and then when it goes down it’s, I sold that months ago right at the highs.

Burry the lede: Michael Burry warns that AI's debt-fueled buildout, rising oil prices near $100, and elevated Treasury yields could strain private equity and credit markets. - Yahoo

Microsoft's Satya Nadella warns AI must fuel real economic growth or "this movie" won't end well, despite $190 billion in spending this year - TNW | Artificial-Intelligence

Microsoft CEO Satya Nadella says trust in US-run infrastructure, which hosts many Chinese open models, will keep American AI competitive. - Business Insider

DRAMatic debut: CXMT, China's largest memory chipmaker, soared 466% on its Shanghai debut to a $487 billion valuation, even as doubts mount elsewhere over AI spending - Fortune

CXMT shares surged as much as 535% on their Shanghai trading debut, briefly making it China's most valuable listed company at about $487 billion. - The Wall Street Journal

China said to be shipping homegrown DUV chipmaking equipment. (not EUV which uses < 1/10 the wavelength) - The Information

China threatens retaliation if the US sanctions its AI firms over claims they copied US models by training on their outputs - Bloomberg

China's state media warns that Beijing's support for open AI models won't extend to unrestricted access to their most powerful capabilities. - Bloomberg

NYU's Scott Galloway likens China's cheap, subsidized AI to its steel-dumping playbook, consolidating market share before extracting margin power later. - Fortune

China is investing multi-trillion dollars in power grids, data centers, and telecom networks to underpin its long-term AI ambitions. - Bloomberg

A US congressional commission delegation could not secure meetings with several leading Chinese tech firms, reflecting Beijing's more cautious approach to AI diplomacy. - South China Morning Post

Pulp fiction: AI firms are bulk-buying rare books through services like ISBNdb, scanning and shredding them, a practice a federal judge ruled fair use - Nitter

Phishing season: CDW survey of 951 IT leaders found 43% of firms suffered AI-enhanced phishing and 37% AI-powered malware attacks in 2026 - ZDNET

Four Minnesota men pleaded guilty to a $2.2 million Medicaid fraud scheme, using ChatGPT to fabricate supporting documentation. - The Cyber Express

Annie Jacobsen warns AI can help design bioweapons even as it speeds outbreak detection, urging AI-powered public health defenses. - the Guardian

AI drug discovery is stalling on missing negative-result data and labs without the interoperable systems needed for full automation - MIT Technology Review

Five architectural patterns give AI agents persistent memory, from short-term working buffers to checkpointed state and tamper-tagged logs of past actions - MachineLearningMastery.com

Intel tested thousands of agentic AI workloads and found agent density and 95th-percentile task latency, not raw agent count, best predict capacity needs - MIT Technology Review

Utility stocks that rallied on AI power demand now face political risk as voters push back against rising electricity costs - MarketWatch

A former NASA robotics chief says orbital AI data centers could work once launch costs fall from $2,900 to under $500 per kilogram - ZDNET

The Navy and DARPA launched the first robotic satellite servicer to extend satellites' lives by six or more years - Naval Research Laboratory

Apply hard: Greenhouse CEO Daniel Chait says job seekers now pay about $20 for AI tools to auto-apply everywhere, driving average applicants per posting up 412%. - Fortune

AI assistants like Gemini, ChatGPT, and Claude now field private research chats, costing brands the SEO and ad-tracking visibility search clicks once gave them, chatbots are sitting on a market research goldmine. - Human-Centered Change and Innovation

Actors are finding AI-generated remakes of their micro-drama roles on apps like MoboReels, with little legal recourse in nonunion contracts. - Business Insider

Face value: Chinese platforms like New Claw and ActID let people license their faces to AI for $15 to $700, fueling a wave of AI microdramas. - Rest of World

Fake AI-generated doctor avatars on Facebook and Instagram pushed a supplement falsely claimed to treat kidney disease, while another was recalled for salmonella. - Yahoo

AI-generated deepfake doctors on TikTok spread health myths in 40% of top videos, UK medical bodies warn of 'huge danger to public safety'. - the Guardian

Meta served roughly 7,600 ads for AI nudify apps from Chinese partner GatherOne, one allegedly linked to child sexual abuse material, watchdog says. - Bloomberg

AI deepfakes clone influencers and celebrities to sell products they never endorsed, part of a wave of AI scams costing Americans hundreds of millions yearly. - New York Post

20% of Americans aged 18-29 have used chatbots for emotional support, and a Stanford study found AI models 49% more sycophantic than humans. - TNW | Artificial-Intelligence

Young adults are letting AI do all their talking for them - The Wall Street Journal

Apple reportedly delayed its AI glasses launch to address privacy issues raised by Meta's AI glasses. - Bloomberg

Meta, Samsung, Amazon, and Plaud are building AI devices that record conversations, raising consent concerns since they're harder to spot than phones. - CNN

MCP for all your appliances: Amazon's Alexa Plus adds Model Context Protocol (MCP) support and a smart home developer toolkit, with Bosch, Whirlpool, iRobot, and Yale Home on board. - Forbes

Models are worse at reviewing their own code. - Greptile

Elon Musk predicts AI will surpass combined human intelligence within five years, forecasting up to a billion humanoid robots. - New York Post

Having a real one, calling Beddoes a race traitor, saying Ilhan Omar must be expelled, calling Chef José Andrès an asshole (seems to have deleted).

Follow the latest AI headlines via druce.ai on Bluesky