TLDR Tech / IT
J-Space: The Logs AI Security Has Never Had
Anthropic researchers reportedly found a small internal workspace in Claude, called the J-space, whose activity can reveal concepts related to the model's reasoning before they appear in its output. Exposing this telemetry would give security teams a third detection layer for AI agents, while acknowledging that the J-lens used to read it is imperfect and currently unavailable to enterprises.
Read the full story