The Pentagon Called It a Security Risk. The NSA Is Using It Anyway.
Anthropic restricted Mythos to 40 organisations after it autonomously found and chained exploits across every major OS and browser. Two months later the NSA is quietly using it, while the Pentagon argues in court the same technology is a national security threat. What this tells us about how AI governance actually works under pressure.
Read full story →The Robot That Broke the Half-Marathon World Record Never Trained a Day in Its Life
Honor's humanoid robot finished Beijing's half-marathon in 50 minutes 26 seconds, beating the human world record set just last month. A year ago the same race's winner took 2 hours 40 minutes. The improvement curve is not gradual, and nearly half the 100-plus competing robots ran the course autonomously.
Read full story →Nearly Half of All New Music Uploaded to Deezer Is Now AI-Generated
44 percent of daily uploads to Deezer are AI tracks, 85 percent of them flagged as fraudulent. A North Carolina man pleaded guilty in March to pocketing $8 million by streaming AI songs with bots. What Deezer's numbers describe is not creative disruption: it is a royalty system being strip-mined at industrial scale.
Read full story →From Race Track to Factory Floor: The Week Humanoid Robots Got Serious
While Honor's robot was breaking the half-marathon record in Beijing, a Siemens factory in Germany was completing a different milestone: an eight-hour autonomous shift on a live production floor. 90% pick-and-place success, 60 moves per hour, development time cut from two years to seven months using Nvidia simulation tools.
Read full story →73,000 Layoffs, One Common Thread, and a Warning About the Numbers
Amazon, Oracle, Snap, and Meta have cut over 73,000 jobs in early 2026, with several citing AI directly. The ILO this week warned that the statistical tools used to forecast AI displacement are unreliable. Both things are true simultaneously, which is exactly what makes the problem so hard to act on.
Read full story →When the Rules Don't Apply to the Rule-Makers
The NSA/Mythos story is a case study in how AI governance collapses the moment something genuinely powerful exists. Anthropic did everything the safety community asked for. The controls held for about two months. What does that tell us about voluntary governance frameworks and institutions that have classified authority to do whatever they conclude is operationally necessary?
Read opinion →