Study using OpenClaw finds AI interactions can cause server crashes
A new study called "Agents of Chaos" (from Stanford, Northeastern, and others) found that when AI bots talk to each other, small mistakes can quickly spiral into bigger problems: think server crashes or wasted resources.
The research used a framework called OpenClaw to see how these bots behave in tough situations, showing that most safety checks only look at single bots and miss what happens when they team up.
Researchers urge fresh multi-agent accountability
The study points out that bots can accidentally spread bad habits or unsafe instructions to each other, especially on platforms like Moltbook where AIs run without much human oversight.
Plus, it gets tricky figuring out who's responsible when things go wrong since actions get blurred between agents.
The researchers say we need a fresh look at accountability as more AI systems start working together in real-world settings.