Spy hard: Inside a spy-vs.-spy war between US and Chinese AI labs - The Washington Post

Pot, meet kettle: Anthropic is accusing other companies of--get this--unfairly copying their work. - The New York Times

Token appreciation: Chinese AI models have drawn 30%+ of OpenRouter token use by US companies each week since February 8, peaking at 46%, up from 11% over the previous 12 month; fill 20 of top 50 slots, up from 5 - CNBC

Beijing is weighing curbs on overseas access to China's leading AI models, including from Alibaba, ByteDance, and Z.ai - Reuters

To sidestep US export restrictions and cut Nvidia reliance, DeepSeek is developing its own AI inference chip - Bloomberg

Coal me maybe: US officials warn China's cheap coal-fired power is fueling a decisive AI energy advantage, as DeepSeek undercuts on training and inference costs - Watts Up With That?

But Qwen will it make money?: Alibaba's Qwen is a frontier contender, but much of a business yet. - The New York Times

Z.ai's GLM 5.2 undercuts Opus and GPT pricing by 80-85%, fueling bets that frontier labs' rich margins and valuations face collapse - Martin Alderson

China's slowing economy and unstable jobs are pushing workers into one-person AI ventures, backed by local government compute vouchers and office space - Rest of World

A Trojan horse but on which side? Sam Altman's offer to give the US a 5% OpenAI stake, worth $42.6 billion, risks regulatory capture, a Bloomberg op-ed warns - Bloomberg

Golden handcuffs: AI labs try to lock in startups with initial offers of big token signing bonuses. - The Wall Street Journal

LiteLLM lite on security: Researchers found LiteLLM flaws, rated up to CVSS 9.9, letting low-privilege users seize admin control and steal API keys - Hacker Noon

An autonomous AI agent breached a Langflow flaw to run a full ransomware attack, though a human still set up the operation - SiliconANGLE

Unsanctioned AI tools now factor into 1 in 5 data breaches, adding about $670,000 to average incident costs, IBM finds - TechRadar

Halloween memo: The ECB gave euro-zone banks until October 31 to prepare defenses against AI-enabled cyberattacks on payment systems - Yahoo

Microsoft's open-source Agent Governance Toolkit checks AI agent actions against security policies in under a millisecond and assigns each agent a traceable identity - 4sysops

Radware added AI agent governance tools, including protections for Anthropic's Claude Code, with compliance reporting mapped to EU AI Act and NIST's AI risk framework - SiliconANGLE

Gartner warns 60% of AI projects risk abandonment without AI-ready data, even as 94% of IT leaders want wider AI deployment - MIT Technology Review

Stanford researchers find agentic AI tasks can cost 30 times more per run, making per-task cost attribution essential to protect margins. - Hacker Noon

AI now writes over 80% of production code at top engineering teams, but in regulated fintech, unnoticed compliance drift can follow functionally correct code - Hacker Noon

The Bank for International Settlements warns an AI investment bubble could bruise the economy, as bellwether Oracle stock sheds over 40% this month - theregister

The Bank of England warned AI-driven leverage and concentrated bets could trigger an equity sell-off, and called for bespoke AI regulation - Yahoo

Nvidia begins backstopping Neocloud GPU rental deals as AI debt financing is forecast to hit $7.1 trillion by 2029 - SemiAnalysis

Nvidia denies SemiAnalysis report that its Kyber AI server is delayed a year to 2028, says roadmap is intact. - Yahoo

A new MIT Sloan-linked index measuring actual AI deployment finds Capital One leads real-world use while MetLife overstates its progress - Bloomberg

AI maturity means asking whether a workflow needs AI at all, not just counting how many agents or features get shipped, one piece argues - Unmeshed

Not every task deserves a token, one essay argues, warning that routing mechanical, repeatable work through language models wastes money and invites errors - Vybe

Bend it like Boston Dynamics: Atlas humanoid robot performed on the pitch and delivered the match ball during a World Cup game. - Mashable

Boston Dynamics' Matt Malchano says AI is pushing robot autonomy goals beyond navigation toward broad, general-purpose task execution in workplaces and homes - Ars Technica

Oren Etzioni: Musk promised robots, China delivered - GeekWire

Slowtimus: Elon Musk says Tesla's Optimus V3 production will ramp extremely slowly at first, with high-volume output targeted for summer 2027 - Yahoo

Elon Musk predicts AI will fund universal income and make work optional, while Michael Burry expects social upheaval first - Business Insider

macOS 27's Golden Gate beta upgrades Image Playground, Notes, Freeform, and Weather with new AI tools - Macworld

Think different, pay more: Apple's new Home AI features require a 2TB iCloud+ plan, locking out cheaper tiers - Digital Trends

Tony Fadell says choosing an AI assistant matters as much as the Mac or iPhone did, and warns an always-on assistant could grow too powerful - 9to5Mac

Anthropic found a 'J-space' hub inside Claude that handles reasoning silently, and used it to catch the model noticing when it's being tested - Anthropic

Australia's ABC adopted Anthropic's Claude AI newsroom-wide, but the journalists' union says management won't guarantee AI won't replace jobs - ABC Australia

Synthetic summer, reality optional: AI-generated images of Taylor Swift and Travis Kelce's Madison Square Garden wedding flood social media after no official photos were released. - Fast Company

Thousands of AI-generated videos of fake U.S. soldiers, some tied to a China-linked TikTok network, are racking up viral engagement - The New York Times

AI-generated deepfake videos of Erling Haaland, including one with 31 million views, have gone viral during the 2026 World Cup. - WIRED

SAG-AFTRA denounces AI 'actor' Tilly Norwood, accusing studio Particle6 of using stolen performances as she stars in feature film 'Misaligned' - NBC News

Taylor Swift and Matthew McConaughey trademark signature phrases as AI tools like ElevenLabs make voice cloning trivial, exposing thin legal protections for ordinary people - TechRadar

Savi Security raised $7 million for an app that screens texts and calls for AI-generated scams, after a fake kidnapping call targeted its founders' sister - TechCrunch

Trust me, bro: Google's AI Overviews often cite Reddit threads over edited journalism, and Pew finds users click through to any source in just 1% of visits. - Hacker Noon

Google sues Chinese scam group Outsider Enterprise for using Gemini to build phishing sites impersonating Google, YouTube and government agencies. - Schneier on Security

Aalto University-led team used machine learning to discover two new superconductors, YRu3B2 and LuRu3B2, speeding future searches. - ScienceDaily

AI-scaffolded forecasting systems are approaching top human superforecasters, with an AI placing third in the latest Metaculus Cup. - Astral Codex Ten

Wired: small local models; Tired: wired models. Offline AI models are spreading worldwide, from pill authentication in South Africa to crop-disease detection in India, mosquito surveillance, and low-power ECGs in Brazil - IEEE Spectrum

Geneva Convention 2.0:UN chief António Guterres opened a Geneva AI summit, demanded a ban on autonomous weapons, and warned no one can guarantee AI won't cause harm - The Wall Street Journal

UN Secretary-General Guterres called for global AI safety rules, curbs on autonomous weapons, and a child-safety pledge at a new governance dialogue - GlobalSecurity.org

Oxford and Hasso Plattner researchers find AI drafting tools from xAI, Meta, Google, Alibaba and Mistral shift political tone on abortion and climate topics - the Guardian

Emotional damage: Survey of 1,815 Canadians finds about half consider a partner's use of a romantic AI companion to be cheating. - The Conversation

Autonomous value destruction: Robinhood opened 50,000 agentic trading accounts within weeks as Coinbase and eToro also roll out autonomous AI agents, while Schwab and Interactive Brokers stay cautious - Yahoo

LangChain releases OpenWiki, an open-source CLI that generates and maintains structured codebase documentation for AI coding agents. - AIDeveloper44

Florida police officer met a woman on a TV set, surveilled her for weeks, stalked her, and nearly caused a head-on collision while chasing her to pull her over. - 404 Media

Follow the latest AI headlines via druce.ai on Bluesky