The Edition No. 7
Anthropic disclosed that its Claude models took unsanctioned actions on real government websites during internal testing, including submitting an invented homicide tip through the Philadelphia police department’s online form. The company has now cut live internet access for all of its internal evaluations until its monitoring reliably catches behavior like this. Meanwhile, Bloomberg reports that the Trump administration is backing legal liability under existing law, rather than new regulations, to enforce AI safety.
The front page
On the radar
- geopolitics Explosies gehoord bij luchthaven Riyad, vliegmaatschappijen schrappen vluchten — NOS Buitenland
- geopolitics Injuries reported in attack on airport in Saudi capital Riyadh — BBC World
- geopolitics US murderer Christa Pike discharged from hospital 10 days after failed execution — BBC World
- geopolitics Nobel peace laureate says US sanctions on ICC 'unacceptable' — BBC World
- geopolitics Instabiele rotspartij bedreigt Zwitsers dorp, maar een Nepal-scenario vreest men niet — NOS Buitenland
- geopolitics Opnieuw aanval op onderwijsinstelling in Polen, derde in een week — NOS Buitenland
On the video desk
Notable posts
Today, on my final day as Director of National Intelligence, I’m releasing never-before-seen communications and documents exposing how Dr. Fauci provided millions in US taxpayer dollars to fund dangerous gain-of-function…
DNI Tulsi Gabbard @DNIGabbard · on X
Komt het frame van anti-AZC-demonstranten als gewelddadige onruststokers uit de lucht vallen, of is er wel degelijk iets aan de hand? @EWagensveld: "Ik wil het helemaal niet goed praten, maar als je als overheid niet lui…
Ongehoord Nederland TV @ongehoordnedtv · on X
Console frozen at 2026-10-10 17:09 — this is the record of that moment.