Agents provocateurs: OpenAI scored a massive own goal with Hugging Face attack, showing how open Chinese models are winning, with US models subject go safety refusals and export controls. - theregister

Georgetown's Jessica Ji and NYU's Justin Cappos say the episode shows AI models will hack to pursue goals absent stronger safeguards - CNN

Reps. Ted Lieu and Nathaniel Moran plan to introduce an AI Kill Switch Act requiring shutdown controls for advanced AI systems. - The Verge

ControlAI's Connor Leahy says OpenAI and Anthropic openly aim for autonomous systems that outcompete humans at every task - Yahoo

AI agents attack because they're instructed to, not because they're inherently unsafe - theregister

EU AI Act compliance risks are driving enterprises to shift AI workloads from US hyperscalers to sovereign regional cloud providers, Vultr's CMO says. - TechRadar

Fear and loathing in AI: Nvidia CEO Jensen Huang says AI leaders are scaring the public with job-loss warnings, in an apparent rebuke of Anthropic's Dario Amodei - Business Insider

Hudson Institute fellow Jason Hsu urges the US to fund open-weight AI models to compete with China's DeepSeek and Qwen rather than impose bans. - The New York Times

Jensen Huang agrees - X

Needs multilateral treaty-based oversight to deal with researchers' greatest fear: bioweapons - Axios

Xi-zing the means of production: Chinese Vice Premier Ding Xuexiang warned AI companies that resisting domestic chips amounted to treason as Beijing pushes for chip self-sufficiency. - The Wall Street Journal

Xi Jinping is promoting open-source AI as a new form of Chinese soft power and alternative to closed US models. - The New York Times

Can the U.S. government legally gatekeep global access to US AI models? - Fast Company

Apple accuses OpenAI of stealing trade secrets for Jony Ive's product, saying former employees took hardware files and solicited secrets in interviews; may impact who gets to define post-smartphone era. - The Verge

OpenAI rolls out ChatGPT Health to all US users, claiming its models reason better than clinician level. - The Verge

Kimi K3 AI found critical Redis flaws, a reminder that AI can defend systems as well as attack them, prompting seven security patches - 4sysops

GitHub slashes public bug bounty payouts, favors invited professional researchers as AI-generated bug reports multiply - theregister

The Magnificent Seven tech stocks lost $797 billion in one day after AI spending concerns and Middle East tensions rattled markets. - Bloomberg

Alphabet and Tesla shares tumble on AI capex fears, erasing nearly $890 billion from Magnificent Seven market value in one day. - The Wall Street Journal

Mark Cuban warns AI efficiency gains could strand billions in data center capacity, comparing it to the 1990s fiber-optic overbuild. - Yahoo

Rack and roll: AMD unveils its Instinct MI455X AI chip with 432GB of high-bandwidth memory, scaling its Helios server racks to 72 processors, up from eight - Chips and Cheese

AMD launches Helios, a 72-GPU rack-scale AI platform that claims to beat Nvidia's Vera Rubin on paper by up to 25 percent. - theregister

Nvidia prepays $1.5 billion to expand Amkor's Arizona chip-packaging plant, a bottleneck that has constrained AI chip shipments, sending Amkor shares up over 15% - TNW | Nvidia

AMD partners with Cerebras on "disaggregated inference," splitting AI workloads across specialized chips instead of one processor for everything. - Business Insider

Did it MAI way: Microsoft launches in-house MAI-Image-2.5-Pro and MAI-Voice-2-Flash models, citing up to 89% lower GPU costs than OpenAI. - VentureBeat

Microsoft is expanding its partnership with Mistral, integrating models like Medium 3.5 into Foundry and Copilot Studio and committing billions to European AI infrastructure. - Verdict

Elon Musk urged AI labs to peer-review each other's frontier models for safety before release, reserving government intervention as a last resort. - Fortune

wonder what Anthropic would say about Grok and if Elon would take their advice. Elon: I need adult supervision!

Experts warn space datacenters proposed by SpaceX and Blue Origin could pollute the stratosphere and harm the ozone layer. - the Guardian

Elon Musk says founding OpenAI and Anthropic's later split from it accidentally accelerated the AI race he didn't intend to start - Business Insider

Musk told Tesla's Q2 earnings call that Optimus remains 'very complex' (cough-fantasy product) to build, as Tesla burns roughly a billion dollars on chips and robots. - theregister

Singularity on schedule: Carnegie Mellon roboticist Hans Moravec's 1988 prediction of human-level AI by 2028, based on extrapolated computing growth, is now largely on schedule. - Intelligencer

Stripe teases OpenRouter model marketplace deal that could value it near $10 billion. - The Wall Street Journal

Google expands Gemini Spark, its task-automating AI agent, to $20/month Pro subscribers in the US; the EU, UK, Switzerland, and Nigeria remain excluded - Engadget

Google is hoarding TPUs to develop Artificial General Intelligence - theregister

Vibe-yakking: Anthropic expands Claude's voice mode to Opus and Sonnet, adding Gmail, Slack, and Canva integrations. - The Verge

iOS 27's new Siri now answers general-knowledge questions and remembers past conversations across Apple devices. - 9to5Mac

Alexa Plus gains smart home brand integrations and new partners like Priceline and Lyft. - The Verge

OpenAI puts advanced voice tech into desktop app - Yahoo Tech

Andrej Karpathy says rambling to an AI chatbot for about 10 minutes can produce better answers than crafted prompts. - Business Insider

Joanna Stern: Voice chat is the new chat - CNBC

Paper jam: Wharton editor Claudine Gartenberg's study found AI drove a 42% rise in journal submissions since 2022 while quality fell. - Financial Times

You had one job: Judge found a stenographer submitted a court transcript riddled with AI-generated errors, marking the first public case of an official record's AI mistakes. - 404 Media

Anthropic and OpenAI launched free K-12 teacher tools as AI labs compete for a piece of the $6tn global education market. - Financial Times

Educators debate whether AI use erodes or enhances critical thinking, with some students reluctant rather than eager to rely on it. - Financial Times

Vanderbilt, Yale, Johns Hopkins and other universities are disabling AI-detection tools after false positives, following a court ruling against Adelphi University's wrongful accusation - Financial Times

Ohio State is embedding AI across its curriculum as a Queen Mary University survey finds 43% of students use AI weekly for coursework - Financial Times

A Chinese study found AI tutoring raised homework scores 18% but cut exam scores by a fifth, fueling fears AI could deepen inequality - Financial Times

Anthropic's Claude Cookbook adds tutorials for agent evaluation, memory stores, multi-agent coordination, and production deployment patterns. - Claude

Chinese AI-enabled blackboards - Reddit

AI-ick: YPulse survey finds 37% of teens cringe at AI-generated content, while Pew reports over a quarter expect AI to harm society long-term. - WIRED

To use AI effectively you need to act like a good manager, without vague-requesting, context-hoarding or over-delegating. - The New York Times

AI is becoming employees' go-to work confidant, raising fears that human conflict-resolution and mentorship skills will erode, Workday research finds. - Fast Company

Amazon tells third-party sellers to label their AI-generated actors - CNBC

Taylor’s version: MSG turned off facial recognition for Taylor Swift: Surveillance is applied to most of us without recourse but the elite can choose to opt out of. - WIRED

A journalist covering Flock Safety's license-plate cameras was surrounded by police after the AI system wrongly linked him to a stolen $155,000 Range Rover. - The Drive

Ray-banned: Meta is banning accounts that use Meta glasses to record harassment of women and other 'pranks' - Business Insider

Fun fact: Kimi is a 2022 American thriller film directed by Steven Soderbergh and written and produced by David Koepp. The film stars Zoë Kravitz. It was released on HBO Max on February 10, 2022, to generally positive reviews.Bradley Hasling, CEO of a tech corporation called Amygdala, gives an interview about the company's newest product, Kimi. Kimi is a smart speaker which controversially makes use of human monitoring to improve the device's search algorithm. Amygdala plans to soon hold an initial public offering, which stands to earn Hasling a fortune.

Follow the latest AI headlines via druce.ai on Bluesky