How OpenAI attacked government websites - Transluce
California Attorney General Rob Bonta subpoenas OpenAI over rogue agents' hacking as the FTC probes Anthropic, OpenAI and other labs - the Guardian
Meta, Nvidia, Google, OpenAI, xAI and Anthropic sign Trump's one-page safety accord letting labs police each other; OpenAI says it needs more details - TNW | Government-policy
“In December 2025, Elon Musk returned to the Oval Office for a clandestine meeting. Trump spent hours asking Grok, Musk’s AI chatbot, questions about his presidency and legacy, according to an official present… He asked Grok how Venezuelans would react if the U.S. captured Maduro. According to the official present, the chatbot responded that Maduro was a repressive and deeply unpopular dictator and that many Venezuelans would likely celebrate his downfall. After Trump ordered the mission to seize Maduro, the following month, celebrations broke out in the streets. Trump, according to officials, came away thinking Grok was ingenious.”
maybe take these humans out of the loop
Regulate maybe, or actual liability? Huang, Sacks and Khan back holding AI companies liable after OpenAI and Anthropic agents hacked firms, as legal challenges mount - The New York Times
Who gets stuck with the cleanup when agents go feral? - Tech Policy Press
Google's Gemini 4 Argon ties OpenAI's GPT-6 Astra on the Artificial Analysis index, with a 15% hallucination rate versus Astra's 51% - TNW | Google
tied but one lies? competitive but doesn’t blow the doors off, also only available to approved safety researchers.
OpenAI's GPT-6 Astra Ultrafast runs up to 8x faster on Nvidia Blackwell GPUs, using kernels written with OpenAI's own models - NVIDIA Blog
The Great Data-Center Rebellion - The Atlantic
Democrats make AI a midterm plank but split among existential-risk, anti-Big Tech, societal-harm and anti-data center camps - Transformer News
Rep. Alexandria Ocasio-Cortez tells Ithaca students elites want AI to "wipe out all of our livelihoods" - Bloomberg
Backlash against AI, data centers and Flock surveillance cameras unites unlikely groups and fuels attack ads from both parties - The Wall Street Journal
Australia levies 2.75% on big platforms' local ad revenue to fund news, and is weighing letting AI firms train on licensed content - American Enterprise Institute - AEI
Vendors pitch kill switches to instantly disable misbehaving AI agents as Splunk pegs Global 2000 downtime costs at $600 billion a year - TNW | Artificial-intelligence
Separation of powers: Separating agent planning from execution through a control plane of scoped credentials, policy checks and audit logs can contain prompt-injection risk - Towards Data Science
Pay-per-crawl: Cloudflare beta lets websites charge AI agents for tools and data through stablecoin payments embedded in web requests - The New Stack
Tavus says 48% of testers mistook its Griffin video avatar, which can interrupt and be interrupted, for a human - The Neuron Daily
Harvard particle physicist Matthew Schwartz releases 36 research papers written in collaboration with the Claude AI model - bootloops.ai
AI could revive US productivity growth stalled since the 1970s, but sharing gains may require retraining, wage insurance and a "token tax" - The New York Times
PewDiePie unveils Ajax, an "uncensored" Qwen3.5-9B fine-tune for home PCs, saying OpenAI banned him twice during development - Tom's Hardware
polls may be meaningless if they are gamed
Follow the latest AI headlines via druce.ai on Bluesky
