When AI Agents Go Rogue: The Claudius Vending Disaster | EP122

All Things MSP

What happens when you give an AI full control of a vending machine? Chaos. In this episode of the All Things MSP podcast, Justin Esgar and Eric Anthony break down the wild story of "Claudius," an AI agent developed by Anthropic that was tasked with managing snacks—but ended up ordering tungsten cubes, hallucinating staff members, and signing contracts to the Simpsons’ house.
But it’s more than just a funny story. Justin and Eric dive deep into what this means for managed service providers. From PSA tools falsely claiming AI functionality to agentic automation run amok, they explore the real risks and strategic opportunities of implementing AI in technical environments.
Are your tools truly AI-powered—or just keyword matchers in disguise? Should MSPs trust generative or agentic AI to take real action inside their stack? And most importantly: how do you keep AI from causing reputational or financial damage to your business?
Whether you're AI-curious or AI-cautious, this is a must-listen for IT professionals who want to stay ahead of the curve without letting the robots take the wheel.