[01]
White House Asks OpenAI, Anthropic to Delay UK AISI Access
The White House asked OpenAI and Anthropic to withhold new models from the UK's AI Security Institute pending US review.
// daily-digest · fri 25 sep 2026 · morning refresh · last 72h focus
Ten items on AI policy and lab economics, Ruby and Rails, AI-era essays, and an astronomy wildcard.
— refreshed for you, in 10 items.
The White House asked OpenAI and Anthropic to withhold new models from the UK's AI Security Institute pending US review.
Anthropic had 201 employees' AI agents negotiate book trades on their behalf to test preference alignment and market behavior.
Groups of Google DeepMind alumni are raising venture funding for new AI startups amid a wave of departures.
Oracle invoked force majeure on its New Mexico Stargate data center after energy permitting delays threatened the 2028 timeline.
OpenAI expanded ChatGPT Ads to Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam, and Taiwan.
Latent Space's daily AI roundup covered new frontier models, agent infrastructure, and a move to Beehiiv.
Rails 8.1.4 shipped with bug fixes and reliability improvements across Active Record and other component gems.
Simon Willison argues MCP still provides useful security and access-control boundaries despite claims that agents made it obsolete.
Scott Alexander examines neuralese, AI models reasoning in internal vector representations instead of readable text.
Astronomers detected the first globular cluster stellar stream outside the Milky Way, offering a new way to probe dark matter.