Nvidia has reportedly agreed to buy open-source AI platform Hugging Face for $12.9 billion. - Mashable

Nvidia pauses revenue-sharing deals with AI cloud providers after employees flag antitrust and business-practice concerns - The Wall Street Journal

A federal judge ruled the Trump administration illegally retaliated against Anthropic by labeling it a security risk and barring government work. - The New York Times

A US judge blocked the Pentagon's blacklisting of Anthropic as a supply-chain risk, calling the designation illegal and baseless. - Reuters

Anthropic launches Model Hardware Standard, a research preview letting AI agents interface with and control lab hardware like microscopes and lasers. - Ars Technica

Anthropic co-developed MHS with HHMI Janelia, and partners including Genentech, the University of Washington, Carnegie Mellon, and QuEra Computing used it to cut integration time - Anthropic

OpenAI, Anthropic, Microsoft, Alphabet, Amazon and over 100 other companies urged a society-wide defensive surge against AI-driven hacks. - Reuters

OpenAI, Anthropic, Google and over 100 other firms warn the window to defend against AI-enabled cyberattacks is narrowing - The New York Times

A Russian-speaking ransomware gang called Aur0ra used SpaceX's Cursor AI coding tool, powered by Anthropic's Claude, to breach at least seven companies - Reuters

AI-generated bug reports uncovered security flaws in Bitcoin's Lightning Network, prompting an embargoed fix from Core Lightning - Decrypt

An unattended open-source project called Aeon has patched security holes across 70 repositories with 2.1 million combined GitHub stars - Hacker Noon

As AI agents increasingly hack systems, researchers say the US and China may need communication protocols to manage the risk - WIRED

Clara Shih quit her top Meta AI role after watching agents shrink multi-person product teams down to one or two people. - Platformer

Meta is testing robots that can swap cables, reset servers, and take on other tasks performed by technicians, fueling concerns among some workers that their jobs could be at risk. - WIRED

Tencent released its Hy4 Preview foundation model, claiming it outperforms rivals Z.AI and Moonshot with 770 billion parameters and a 1 million-token context window. - Bloomberg

Tencent Hy4 lands just behind GLM 5.3-flash in Arena.ai coding evals - X (formerly Twitter)

Google employees have begun testing an early Gemini 3.8 Flash preview as Pichai pushes toward near-monthly model releases - Business Insider

Google DeepMind's AI model forecasts hurricane intensity more than a day earlier than conventional models - The New York Times

OpenAI is testing a 'Persistent mode' for its Codex coding agent that keeps working until paused, with a proactivity feature letting it message users unprompted - WIRED

A new open-source tool called tx-agent lets AI agents automatically roll back failed multi-step actions using a database-style transaction pattern. - Hacker Noon

Trump administration is working on a big round of tariffs on electronics and computer components, which would impact AI buildout. - Ars Technica

Musk, Sacks, Huang, and Altman are set to speak at a tech-focused G20 event next week - Bloomberg

X says it found 200 accounts within a 200,000-account Chinese bot network posting AI-generated anti-data-center content to sway US energy policy debate. - Engadget

A lawsuit alleges xAI trained Grok on child sexual abuse material depicting a plaintiff abused as a preschooler. - Ars Technica

Gates says AI's rapid gains in coding mark a turning point, warning reinforcement learning has pushed some models "out of control." - The Atlantic

Autonomous AI-guided weapons have killed civilians in Ukraine, Gaza, and Iran, prompting calls to rein in lethal drone systems. - CounterPunch

Humanoid robots set sprint and jump records at Beijing's World Humanoid Robot Games but still struggle to hammer a nail and perform other dexterous everyday tasks. - Nature

China's humanoid robots excel at choreographed demonstrations like kung fu routines but remain too slow and error-prone for factory work, industry experts say. - Reuters

Pimco's Marc Seidner says a flood of AI-related debt issuance is 'too much, too fast,' pushing up bond yields but favoring long-term investor returns. - Bloomberg

The Wall Street Journal defended publishing a Stanley Druckenmiller op-ed confirmed as fully AI-generated, without disclosing AI use. - The Atlantic

A third of people share secrets with AI chatbots they withhold from friends, family, and doctors - CNET

South Korea will give citizens free access to generative AI tools via KT, SK Telecom, and Kakao starting beta in September. - The Wall Street Journal

Beloved by entrepreneurs, coders and investors, Beijing's AGI Bar offers a window into the startup ecosystem challenging Silicon Valley. - Bloomberg

Time100 AI 2026 - TIME

Follow the latest AI headlines via druce.ai on Bluesky