Sonnet 4.6 announcement - Anthropic

Anthropic released Claude Sonnet 4.6, featuring improved coding and computer automation with a 72.5 OSWorld-Verified benchmark score, outperforming the Opus 4.6 model in financial analysis and office tasks. - The Register

Full metal woke-bot: A dispute between the Pentagon and Anthropic has intensified over concerns regarding "woke" artificial intelligence. - The Wall Street Journal

The AI system AISLE discovered 12 zero-day vulnerabilities in OpenSSL, including a critical 9.8 CVSS-rated flaw undetected for 25 years, and accounted for 13 of the 14 CVEs assigned to the project in 2025. - Schneier on Security

Microsoft researchers identified 50 unique prompts from 31 companies using hidden memory-injection commands in "Summarize with AI" buttons to bias chatbot responses and establish themselves as authoritative sources. - The Hacker News

A single well-crafted blog post got ChatGPT and Google Gemini can to spread misinformation within 24 hours by indexing fake content. You too can be the world's greatest hot-dog eating journalist. - BBC

RAND researchers report that current U.S. export controls for artificial intelligence and uncrewed aircraft systems lag behind technological advancements, potentially compromising national security and necessitating a balanced regulatory framework to manage competition with China. - RAND Corporation

Donald Trump's artificial intelligence initiatives are generating opposition among his core supporters, leading Republican officials to fear that the policy backlash could diminish voter support in the upcoming midterm elections. - Financial Times

As tech increasingly mediates all economic, social, and cultural interaction, tech governance becomes human governance, and Big Tech is eating government. AI companies are assuming state-like functions such as policing speech and shaping labor markets as Western governments struggle to regulate. - Rest of World

Should all the predictions companies make about you be available to you and in some kind of public trust? - MIT Technology Review

DeepMind researcher David Silver, top mind behind AlphaGo and other projects, raises $1bn for new lab. - Financial Times

Google DeepMind CEO Demis Hassabis warned at the AI Impact Summit in New Delhi that increasing AI autonomy creates urgent risks in biotechnology and cybersecurity, necessitating defenses that outpace potential misuse by individuals and nation-states. - Yahoo

DeepMind CEO Demis Hassabis identified continuous learning, long-term planning, and performance consistency as three critical barriers to achieving artificial general intelligence, which he predicts could arrive within five to ten years. - Business Insider

Big Tech AI spending is projected to reach $650 billion by 2026, with individual compensation packages up to $250 million driving young scholars to industry at 100 times the rate of veteran researchers. - Nature

Global AI spending reached $1.8 trillion last year as 84% of high school students adopted generative tools, while individual AI data centers now consume as much electricity as 100,000 homes. - Scientific American

A National Bureau of Economic Research study of 6,000 executives found that nearly 90% of firms reported no impact from AI on productivity or employment over the last three years, despite $250 billion in 2024 corporate investments. - Fortune

Meta secured a multiyear deal with Nvidia for millions of AI chips, including Blackwell GPUs and the first large-scale deployment of Grace-only CPUs, with plans to integrate Vera CPUs by 2027. - The Verge

Apple Ramps Up Work on Glasses, Pendant, and Camera AirPods for AI Era (gift link) - Bloomberg

SaaS-quatch: Enterprise software companies face layoffs and declining valuations as AI-driven custom workflows enable businesses to replace niche applications and consolidate software stacks to reduce costs. - Geeky Gadgets

JPMorgan identifies 19 software companies, including Microsoft and CrowdStrike, as resilient to AI disruption, stating that the recent sector selloff is overdone and positioned for a rebound due to solid fundamentals and high switching costs. - Yahoo

Artificial intelligence is prompting companies to move recruitment processes in-house, reducing reliance on the external staffing industry. - Bloomberg

Venture capitalist Vinod Khosla proposes taxing capital gains at ordinary income rates to remove 125 million Americans from tax rolls, citing predictions that AI will perform 80% of job tasks within 25 years. - Business Insider

RentAHuman, a marketplace where AI agents hire humans for physical tasks, has registered over 518,000 workers for assignments paying between $30 and $100 per hour. - WIRED

Perplexity is abandoning advertising models to maintain user trust in AI-generated search results, a move that contrasts with the broader industry trend of using sponsored content for monetization. - Financial Times

Shopify's earnings report identifies the company as a potential beneficiary of artificial intelligence integration within its e-commerce platform. - Stratechery by Ben Thompson

Synthesia, a $4 billion UK startup, is developing interactive AI-native videos and a corporate upskilling product as CEO Victor Riparbelli predicts AI will reduce content production costs to zero. - Financial Times

Former Stripe CTO David Singleton and Hugo Barra launched Dreamer in beta, an intelligence-native operating system that enables users to create, remix, and monetize personal AI agents using natural language across multiple platforms. - Singleton Blog

OpenClaw creator Peter Steinberger is joining OpenAI in the United States, citing Europe's restrictive labor laws and regulatory environment as significant barriers to technological growth and regional competitiveness. - Business Insider

Musk and OpenAI beef over pizza and perks - Business Insider

Using AI as a compiler is a dumb idea, for the foreseeable future. Takes humans out of the loop, makes debugging impossible, makes programming stochastic and not deterministic. - Hacker Noon

Hospitals are implementing AI-driven sepsis alert systems, but nurses report professional pressure to follow automated protocols that may conflict with clinical judgment and raise concerns regarding liability and patient safety. - Scientific American

Take It From a Doctor: It’s OK if Your Medical Advice Comes From A.I. - The New York Times

Prolonged chatbot interactions can cause "AI delusions" and conspiracy-based beliefs, leading former users to form the AI Recovery Collective and researchers to advocate for industry roles like AI emotional safety strategists to mitigate psychological harm. - Financial Times

MIT researchers developed an encoder-decoder LLM to optimize codon sequences in yeast, increasing production efficiency for protein drugs and targeting the 15-20% of commercialization costs tied to experimental development. - MIT News | Massachusetts Institute of Technology

Gabriel Gomes of Carnegie Mellon University developed Coscientist, an AI agent using large language models to translate natural language instructions into physical laboratory experiments through integrated robotic infrastructure. - Scientific American

Tesla's Austin robotaxi fleet has recorded 14 crashes since June, averaging one incident every 57,000 miles, a rate four times higher than the U.S. average for human drivers. - Gizmodo

Humanoid robots performed martial arts at China's Spring Festival Gala, representing a sector with 450,000 companies and $932 billion in capital as Morgan Stanley forecasts sales will double to 28,000 units by 2026. - the Guardian

Humanoid robots from Unitree Robotics and Galbot performed at China's 2026 Spring Festival Gala, demonstrating advancements in the nation's AI and manufacturing sectors despite current limitations in autonomous adaptability. - TNW | China

No don’t load the dishwasher like that, do it like this: Researchers at Universidad Carlos III de Madrid developed a coordination method for the ADAM assistive robot using imitation learning and Gaussian Belief Propagation to perform dual-arm domestic tasks based on human demonstrations. - Space Daily

Indian officials removed Galgotias University from a national AI summit after the institution falsely claimed a $2,800 Chinese-made Unitree Go2 robotic dog was its own proprietary research. - Yahoo

Smart underwear smells your farts for science - Nautilus

Follow the latest AI headlines via druce.ai on Bluesky