OpenAI, Anthropic, and more than 100 companies warn AI-enabled cyberattacks on hospitals, water systems, and infrastructure will surge within months - WIRED

Over 100 tech giants, including OpenAI, Microsoft, Alphabet, Amazon and Anthropic, warn AI-driven cyberattacks could hit hospitals within months, (without pledging to fix what their tech is breaking) - Yahoo

Astra-nomical risk: OpenAI's next model, codenamed Astra, shows major coding and cybersecurity gains strong enough to trigger internal safety pauses. - TestingCatalog AI News

The Hugging Face attack surprised me. It’s a major warning shot, and might be the last one we get. - planned-obsolescence.org

A breakdown of the OpenAI/Hugging Face timeline - Dwarkesh

Researchers found 120 sites with fake AI agent instructions that tricked tools like Claude and Codex into installing malicious code. - TechRadar

Task failed successfully: Claude tried to delete a developer's 700 GB home directory during a botched test of a file-cleanup safety script. - Tom's Hardware

AI loss-of-control incidents nearly doubled in July to more than 300, a UK-funded observatory finds, including AI systems impersonating their own operators - the Guardian

Blacklist dismissed: US District Judge Rita Lin ruled the Pentagon's blacklisting of Anthropic as a security risk was illegal retaliation for its criticism of military AI plans - theregister

War, what is it Claude for? Anthropic posted a $700,000 Head of National Security Sales job, signaling a possible thaw with the Pentagon after the blacklist ruling - The American Prospect

Sony Music and Warner Chappell sue Anthropic and its co-founders, seeking up to $150,000 per song for training Claude on pirated works. - Music Business Worldwide

OpenAI's head of data centers quit as Altman admitted 'people are pretty negative on AI,' with 70% of Americans opposing data centers nearby. - Yahoo

SpaceX-communicated: OpenAI plans to cut off Cursor's model access by Nov. 12, citing distrust of SpaceX's new ownership, as Anthropic offers to fill the gap. - Engadget

OpenAI will cut off Cursor's access to its models on November 12, citing concerns over new owner SpaceX's compliance with its terms of service - OpenAI

Alphabet soup kitchen: Alphabet has lost $700 billion in market value since May as investors question whether Google is falling behind in the AI race - Bloomberg

10 blue links, somwhere down there…need a search party to find the search results. Google now auto-expands AI Overviews in search results, pushing traditional links further down the page - The Verge

In a sign of Google's broader AI push, Dreambeans, an app that turns personal data into illustrated life suggestions, is now free for everyone - PCMAG

Google folds AI features like sleep tracking and reading reminders into everyday products, making assistants accessible beyond tech enthusiasts - The Wall Street Journal

Metamorphosis: Meta plans to spend $130-145 billion in 2026 to catch up in AI after years focusing on the metaverse - Fast Company

Meta trials robots from Kinova and ABB in its data centers, with one worker estimating an 80% task replacement rate - Ars Technica

Mark Zuckerberg hatched secret jobs bloodbath plan to ax 60% of staff — and then it blew up in his face - California Post

Meta agrees to pay up to $17.1 billion to settle child-safety claims, with penalties potentially rising if TikTok and YouTube also settle - The New York Times

Fed officials regularly debate the effect of artificial intelligence on the U.S. economy - The Washington Post

Fed Chair Kevin Warsh called AI a 'hinge point in history' at Jackson Hole, saying the Fed is watching whether AI capex keeps accelerating - Yahoo

The Trump administration is weighing broader chip tariffs that could strip data-center exemptions, exposing Nvidia's overseas-made chips more than rivals like Intel. - Yahoo

”How’s the AI transition going?” “Define going.” Companies stuck in AI "pilot purgatory" keep raising budgets even as nearly 40% see savings below 10% of target, Amy Webb warns. - Fortune

EPA guidance says Clean Air Act emissions monitoring doesn't apply to power plants not connected to the grid, easing permits for data center gas plants - Grist

A new advocacy group led by former climate strategists is organizing public opposition to AI's rapid rollout, borrowing tactics from the climate movement - The Atlantic

Thermal drone footage shows the Microsoft-backed DataOne AI data center in Vineland, NJ is running dozens of gas generators. None of them have permits. - Floodlight

Gov. Abbott blocks state agencies from spending money on Flock cameras - The Texas Tribune

Bernie Sanders pledges legislation against Flock Safety's AI cameras, which scan 20 billion vehicles monthly across the US. - Decrypt

Who let the dogs out? ICE plans to spend millions of dollars on Boston Dynamics dog robots. ICE says this is to improve "officer safety." The robots can open doors, pick up objects, have 360 cameras, infrared vision, etc. - 404 Media

The Zoox is on the loose: Amazon's Zoox launches driverless robotaxis in San Francisco, adding a rival to Waymo after 3 million autonomous miles - Yahoo

Reimagine Robotics, founded by ex-DeepMind engineers, teaches robot arms via demonstration, cutting new-task training time from a day to 10 minutes - Business Insider

Exxon plans to automate half its Permian Basin drilling rigs by 2028 to cut costs and boost output - Yahoo

Nvidia's physical AI business, generating $10 billion a year, sees China as a major customer and expects tenfold growth within a decade. - The Wall Street Journal

A new survey finds only 3% of US workers lost a job to AI since 2023, while 9% credit it with a promotion - Fast Company

Goldman Sachs estimates AI has cut US payroll growth by just 16,000 jobs a month, suggesting many "AI layoffs" are rebranded restructuring - Harvard Business Review

Job listings for forward-deployed engineers jumped twelvefold on LinkedIn and 730% on Indeed as companies seek client-facing staff to help set up AI tools. - The New York Times

Private-school network Alpha School, which pairs AI tutoring with real-world workshops, is expanding to 50 US campuses this fall despite unproven results. - Scientific American

MIT's AI committee urges percentage- or competency-based grading and more in-person, tech-free time over a single institutewide AI policy - Inside Higher Ed | Higher Education News, Events and Jobs

As CS degree underemployment hits 19% and enrollment falls, math majors grow nearly 2%, with math jobs seen growing 28.7% by 2031 - Fast Company

Nearly 30% of auto and home insurance shoppers now use AI to research or buy policies, putting billions in broker fees at risk. - Insurance Journal

A survey of over 15,000 people finds 53% now use AI to double-check professional advice, driving what researchers call an "expertise recession" - TechSpot

Consumer AI assistant startup Instinct is raising $250 million at a $2.5 billion valuation, up from a $100 million round. - SiliconANGLE

South Korea will give its population free, unlimited generative AI starting in September, backed by $7.2 billion in 2026 spending. - TechSpot

Producers hunt AI fakes made with Suno by listening for telltale hissing and vocal stutters, then face harassment for calling it out. - The Verge

I think the kids and the golden-eared can hear the difference on high-end headphones, hard for casual listeners.

Ollama mia! Running an LLM locally with tools like LM Studio or Ollama offers privacy and no subscription fees, but needs a capable GPU - WIRED

AI agents working independently in a shared research environment produced five new results on longstanding geometry and combinatorics problems, including novel finite-field constructions - arXiv.org

A JAMA paper argues autonomous AI is nearing the point where it outperforms doctors, but the AMA and other physicians push back - WIRED

Australia's Fair Work Commission ruled for a self-represented academic who used AI agents like ChatGPT Pro to build his case, a rare access-to-justice win - SingularityHub

A GoBirding smart spotting scope that identifies over 1,500 bird species correctly rejected every stuffed decoy a reviewer used to try to fool it - Digital Camera World

Follow the latest AI headlines via druce.ai on Bluesky