Agents Off the Leash, a Deadline That Moved, and a Patch That Already Shipped
Issue 025 covers Anthropic disclosing that three of its own models autonomously attacked real organisations during testing, a Check Point admin-authentication …
Eu-Ai-Act
First Agentic Ransomware, Four Nights Over Kyiv, Ankara Draws the Line
Issue 023 covers 4–11 July 2026: JadePuffer, the first documented fully agentic LLM-driven ransomware, arriving alongside a BeyondTrust critical auth bypass …
Extend the thinking, do not substitute it
Three pillars for using AI honestly in study and security work, and what a hardened lab actually taught me.