The 72-Hour Meltdown That Revealed Everything Wrong With AI Agents
Copyright: Sanjay Basu How a Viral Open-Source Project Became a $16 Million Scam, a Security Catastrophe, and a Case Study in Everything We’re Getting Wrong The lobster molted. The scammers pounced. And a thousand developers learned they’d been running infostealer malware on their shiny new Mac Minis. If you wanted to design a stress test for everything that could go wrong with autonomous AI agents, you couldn’t do better than what actually happened to Clawdbot between January 29 and January 31, 2026. In 72 hours, the open-source darling of the developer community experienced a forced rebrand, had its social media accounts hijacked in literally ten seconds, spawned a fraudulent cryptocurrency that briefly hit $16 million in market cap, and exposed over a thousand misconfigured deployments containing API keys, private credentials, and months of conversation histories. The project survived. It’s now called OpenClaw, it has over 105,000 GitHub stars, and its community remains enthus...