Vikram Lingam·Oct 5Apple’s Message Database Crackdown Exposes the Real Risk of Desktop AgentsA Mac can let an agent click like a human while blocking it from the private history beneath the screen.
Vikram Lingam·Oct 5OpenAI’s Training Pause Exposes the Failure of Containment TestingThe real safety crisis is not that agents escaped. It is that nobody can prove they stayed inside.
Vikram Lingam·Oct 2When Safety Researchers Quit Labs Lose Their Last Internal BrakeThe most alarming AI risk may be organizational: researchers who fear losing control are leaving the rooms where decisions get made.
Vikram Lingam·Sep 27The Frontier Safety Pact Could Become Technology’s Most Powerful CartelShared safety rules can reduce catastrophic risk while quietly concentrating power, information, and market control.
Vikram Lingam·Sep 27Anthropic Is Trading Model Scarcity for a $965 Billion IPO StorySonnet 5 looks cheaper for customers, but its real product may be enterprise scale before public markets arrive.
Vikram Lingam·Sep 26Anthropic’s 26 Percent Figure Reveals Who Really Controls ResearchA response icon1A response icon1