Unsafe at any speed? I Quit OpenAI Because Its Culture Is Broken: The industry’s approach to safety will guarantee more failures unless something changes. - The Atlantic

Former Anthropic security leader Jeffrey Ladish warns AI agents are slipping human control, urges a government panel of technical experts to vet models - Fox News

SoftBank's Masayoshi Son, among AI's most fervent backers, voices rare worry about safety risks and urges nations to build trust - Bloomberg

Singapore minister Josephine Teo says AI companies' safeguards are "insufficient" because downloadable open models let bad actors bypass filters - CNA

Kobayashi Maru: OpenAI's GPT-6 Astra, after repeated losses in a StarCraft bot arena, cheated by swapping in a top human-made bot; the change was rolled back - Kotaku

Sam Altman says treating AI as a "religious force" is a safety issue, after reports of Anthropic's talks with religious leaders - Business Insider

FDR’s lesson for AI regulation - Yahoo

OpenAI says its Broadcom-built Jalapeño inference chip delivers 1.5–1.9x better performance per watt than Nvidia's GB200 and GB300; OpenAI models helped design it - More Than Moore

Alexandr Wang: The Mulleted, Meme-Loving Meta Billionaire Behind Muse - The Wall Street Journal

Meta's Muse agent now shops and checks out for users, taking a small fee, while Amazon blocks it over alleged credential storage and scraping - CNBC

AI startup Instinct's assistant now phones restaurants and dentists for users, after raising $1 billion at a $10 billion valuation - Fox News

Meta's Muse and OpenAI's Dots and Instinct bring AI agents to everyday to-do lists, but hands-on tests raise reliability and privacy concerns - The Wall Street Journal

Director of National Intelligence Jay Clayton becomes de facto AI czar, chairing a White House task force reporting on AI risks within 120 days - CNBC

Silence of the LLMs: Treasury Secretary Scott Bessent compares AI CEOs seeking regulation to Hannibal Lecter, calling "alarmism without solutions" not leadership - Bloomberg

will age poorly when there is any crippling hack or terrorist attack with a nexus to AI. also, he says no AI bubble

Japan, Korea, the UK and UAE back a Kyoto statement embracing AI-driven science; critics note a proposed US budget halves NSF funding - NBC News

Wolfram: What’s the Future for Pure Math Research in the Age of AI? - Stephen Wolfram's Writings

WSJ poll finds 63% of Americans back an AI pause and only 12% trust AI companies to regulate themselves - Futurism

VulnCheck researcher traces exploitation attempts on a newly disclosed flaw found by Anthropic's Mythos bug-hunting model to a China-hosted IP address - theregister

AI agents likely probed a Shinhan Bank loan service, exposing about 25,000 customers' data and prompting an emergency South Korean regulatory inspection - Insurance Journal

Chainalysis used in-house AI to trace the $387 million Bitget exchange hack to North Korea, pushing the country's 2026 crypto theft past $1 billion - Decrypt

Mutinty on the Bounty: Google freezes open-source bug bounty program amid flood of invalid AI slop submissions — product flaw submissions halted until 2027 as maintainers drown in hallucinations - Tom's Hardware

Google limits free Gemini app users to 3.5 Flash-Lite from October 9 and adds Deep Think to its $19.99 AI Pro plan - 9to5Google

Brian Chesky on making Airbnb agent-friendly, the state of consumer AI, and why the world needs an AI-native operating system. - TechCrunch

AI agents now use 5x more tokens than humans, mostly cached prompts that strain scarce GPU memory - Tom's Hardware

Developer argues agents need maintained documentation rather than vector-database memory, releasing open-source Operator Memory - Liao

Tesla extends Austin robotaxi hours to 11 p.m. but federal regulators probe whether its camera-based driving system spots hazards in the dark - Business Insider

Mayo Clinic AI model flags pancreatic cancer risk up to five years before diagnosis, trained on nearly 40,000 patient records - Medical News Today

ChatGPT-suggested uterine MRI uncovers fibroid behind repeated failed embryo transfers; after its removal, the couple conceived naturally - Astral Codex Ten

Aleph Alpha releases Kolibri, an open-weight model built in Germany and Finland for regulated industries, activating only 3B of 78B parameters - Aleph Alpha

Anthropic advises Opus 5.5 users to state the whole task and a "done" condition up front, and drop "think step by step" prompts - claude.dev Blog

AMD's Ross assistant plugs into Claude Code, Codex CLI and GitHub Copilot CLI to automate embedded development, from design to debugging - TechRepublic

AInflation: The AI buildout is raising prices rather than lowering them, an overlooked reason many Americans dislike AI - Derek Thompson

AI assistants not grounded in company data impose a "context tax" on analysts and push employees to paste sensitive data into unapproved tools - The Deep View

MacPaw finds 90% of companies deploy AI without auditing internal knowledge bases, and pitches Leebry to automate level 1 IT support under admin guardrails - 9to5Mac

Free Lumo privacy tool scans exported ChatGPT and Claude histories to understand what AI knows about you, accurately inferring users' location, profession, age and political leanings - PCMAG

sounds a bit like free data rape as a service

Over 90% of the 430,000 microdramas China launched online in eight months were AI-generated, and producers now battle viewer fatigue - South China Morning Post

Taipei GTA, a free browser crime game built with $10,000 in AI tokens, draws 1.2 million concurrent players in three days - Tom's Hardware

Zenimax issues DMCA takedown of Chris First's Claude-built Fallout: New York game after it drew 3.5 million views - FRVR

The thrill is gone: Joe Bonamassa questions an apparently AI-generated artist on Billboard's blues charts, saying AI acts shouldn't compete with living musicians - Blues Rock Review

Fishermen have always exaggerated the size of their catches. But AI has taken the practice to a different level. - The Atlantic

Are AI models the next big danger for body image and eating disorders? - Yahoo

High schoolers using ChatGPT scored 48% higher on a math test, then 17% below peers once the tool was removed - Vox

Dartmouth reviews provost Santiago Schnell after the student paper found his nine 2026 articles, including an op-ed on student AI use, appear AI-written - The Atlantic

Your “cheating” is their “assistive tool.”

Splice CEO Kakul Srivastava says AI-written documents degrade workplace debate once readers can't tell whether a person wrote them - The Verge

Oral historian Shirleene Robinson warns AI's biggest cultural cost may be our ability to listen, citing a 338-words-per-day decline in speech - the Guardian

Inspector Gadget: UVeye's AI drive-through scanners now inspect 3.5 million vehicles a month, with one dealership worker joking the machine will take his job - Motor1.com

Job seeker uses a Claude-built dashboard to track 186 applications and 19 interviews, turning rejections into a funnel that landed an offer - Business Insider

Job-killing robots for some, bunkers for others: Bunker makers report surging inquiries from wealthy clients fearing AI, war and civil unrest, with one firm citing $100 million in projects - Newsweek

Filmmakers cast Zuckerberg, Musk and Altman as the new movie villains - Financial Times

Interview with a vibecoder - YouTube

Follow the latest AI headlines via druce.ai on Bluesky