Day-Zero: Agents in the Wild

Autonomous agents are moving into real work faster than our ability to measure what happens when they're under attack. Day-Zero is a live adversarial research exercise that puts that question in a room. For one night in Oakland, red and blue teams go head-to-head against a purpose-built multi-agent system. The goal isn't to find a single bug; it's to learn how agents behave under pressure: whether they can be pushed into doing something they shouldn't, whether they keep doing their job while compromised, and what changes when an agent knows an attack is coming. Findings feed an open field report for the wider community. This is the third experiment in the Adversarial AI Experiment Series, following The Penthouse Heist (FR-001) and The Infrastructure Strikes Back (FR-002) . It brings…