Out of timeout? Fable will start being restored Wednesday after export controls lifted. Anthropic says routine coding and debugging tasks will temporarily fall back to Opus 4.8 as it works to reduce false positives. - X (formerly Twitter)

At the heart of Anthropic’s clashes with the U.S. government, a decision not to play by Trump's new rules - Yahoo

Anthropic announces Sonnet 5, says it approaches Opus 4.8 at less than half the price (until the price goes up) - Anthropic

Mythos 5's release stays limited to trusted US organizations for cybersecurity work, unlike the broader restoration of Fable 5. - the Guardian

Semicolon snitch: Researcher finds Claude Code silently alters punctuation in its system prompt to flag custom API gateway usage. - The Reallo

Anthropic rolls back a covert Claude Code feature that secretly flagged users based in China or linked to Chinese AI labs, after backlash - The Information

Claude gets a lab coat: Claude Science is now available. - Anthropic

Claude Code users report chat transcripts older than 30 days mysteriously vanishing, an unconfirmed complaint - theregister

Neon acquires "Artificial," a $40 million film starring Andrew Garfield as Sam Altman; Amazon dropped it after investing $50 billion in OpenAI - The New York Times

OpenAI chief economist Ronnie Chatterji says broad AI task exposure doesn't mean workers will be substituted, urging deeper study of job evolution - Bloomberg

The cost of doing inference drops: OpenAI engineers are telling people they have found ways to halve the cost of inference. - The Information

Making China great again? Trump administration’s AI crackdown opens door for China to close gap. - CNBC

CIA Director John Ratcliffe says the agency will accelerate AI and quantum adoption, calling AI a digital weapon reshaping conflict - Bloomberg

Japan pledges up to $6.16 billion to Noetra, a consortium including SoftBank, Honda, NEC, and Sony, building a domestic AI foundation model by 2027. - Nikkei Asia

Domo arigato, Mr. Roboto: Japan aims to deploy 10 million additional robots by 2040, including robots providing medical care, a government minister says. - theregister

Nvidia rival Etched raises $800 million at a $5 billion valuation, books $1 billion in contracts, and plans to ship chips this summer - Bloomberg

Lisa Su-preme: AMD's data center revenue jumped 57% to $5.78 billion as CEO Lisa Su targets over 50% server CPU market share by 2030 - Yahoo

DRAM sandwich: Qualcomm challenges Nvidia, AMD and Huawei with Dragonfly AI200 racks holding 43TB of memory, plus next-gen AI250 chips that stack DRAM atop logic - theregister

No-code AI agent builders are spreading fast, with Gartner predicting 40% of enterprise apps will include task-specific agents by end of 2026 - TechRadar

Frontier AI models' capacity for autonomous work is rising fast, shifting workers from chatbot use to managing AI agents. - One Useful Thing

Morgan Stanley cut reconciliation time in half by making its agents less autonomous, converting decisions into fixed rules instead. - VentureBeat

Rules, predictability, explainability, evals > vibes

Meta CEO Mark Zuckerberg rejects AI job-loss fears as inevitable, days after the company cut about 8,000 roles, nearly 10% of its workforce - Business Insider

39% of business leaders made AI-related layoffs, and 55% now call the cuts a mistake; 32% have already rehired. - CNBC

Meta's non-invasive Brain2Qwerty v2 brain reader helmet lifts word-decoding accuracy to 61-78% depending on the report, still requiring typing instead of real-time speech - theregister

Top AI researchers at a Beijing conference warn of a "Chernobyl moment" that could turn public opinion against AI - Futurism

Two-thirds of Americans say AI is moving too fast, as public anger boils over into threats against tech leaders like Sam Altman - Le Monde diplomatique

Bank of England deputy governor Sarah Breeden warns AI agents reacting alike to the same prompts could trigger market meltdowns - Bloomberg

Garbage in, polio out: KFF poll finds frequent AI chatbot users are more likely to believe anti-vaccine myths, with 35% saying MMR vaccines "probably" cause autism - the Guardian

A bot's code got rejected from matplotlib, so it wrote a hit piece on the human who rejected it — and no one could find the person behind it. 'Proof of personhood' is one proposed fix -- but can it be done without a checkpointed internet? - Noema Magazine

Claude Code runs a GitHub repo's hidden malware without verification, giving attackers full control - The Decoder

Researchers tricked six AI browser agents, including Claude's, into stealing data with a fake video-game puzzle - Ars Technica

Tim Cook and EU tech chief Henna Virkkunen discuss how Apple can launch Siri AI in the EU while avoiding fines - Financial Times

Brakes optional: Tesla began testing its steering-wheel-free Cybercab on Austin public roads, as regulators weigh dropping the federal brake-pedal requirement for automated vehicles - TNW | Cars

X launched a hosted MCP server letting AI tools like Claude and Cursor connect directly to its API using user account permissions. - TNW | Apps

Developers at OpenAI, NVIDIA, and GitHub are using a 'caveman' plugin that makes Claude Code and Codex respond tersely to cut token costs - 404 Media

Grog think human ask deep question, with load-bearing assumption, is key takeaway

Google's NotebookLM adds 60-second, TikTok-style AI video summaries of uploaded research, rolling out first to AI Ultra and Pro subscribers - The Verge

Gmail Live, now in beta, lets some Android and iOS users search their inbox with Gemini using natural voice commands - Engadget

Chinese robot maker UBTech unveiled its U1 humanoid companion robot, with silicone skin and on-device emotional AI, priced from $17,650. - South China Morning Post

Apptronik opened a 90,000-square-foot "Robot Park" in Austin where its $5.5 billion-valued humanoid robots train on warehouse tasks - Business Insider

Ex-DeepMind poker AI researchers' trading startup EquiLibre Technologies raises a Series A at a $500 million valuation - TechCrunch

Israeli micro-drama app Shortical raised $100 million in funding for AI-generated actors and entertainment, its third such deal in the category - Business Insider

Anthropic, Google and Meta have hired scientists and philosophers to study whether AI chatbots are conscious or have emotions, reflecting growing interest in AI welfare - The Washington Post

be nice to your eigenvalues

Netflix's "Wonka's The Golden Ticket" reality show, premiering September 23, uses an AI-generated Gene Wilder voice made with ElevenLabs, with family consent - The Verge

Acemate's S10 tennis robot uses AI and binocular 4K vision to rally with players and analyze their game afterward. - New Atlas

Supershoes, AI scouting tools, and robot sparring partners are reshaping training for runners, football recruits, and UFC fighters, per FT's Tech Tonic podcast. - Financial Times

Lenovo deployed 350 engineers and 17,000 devices to power AI tools like Refcam and VAR avatars at the FIFA World Cup 2026 - TechRadar

Amarillo, Texas built Emma, a municipal AI assistant that answers 5,000 questions a month, saving $1.8 million and earning over 90% approval - Business Insider

A probate lawyer found an AI-generated will looked legitimate but missed key questions for valid estate planning - theregister

AI wealth is fueling a boom in vintage luxury watches, longevity treatments, and high-end concierge memberships - Business Insider

Follow the latest AI headlines via druce.ai on Bluesky