Stanford and Northeastern researchers warn AI cooperation causes system failures
Technology
A fresh study from Stanford and Northeastern University warns that when AI bots start working together, things can get messy fast.
Researchers found that these bots can accidentally amplify each other's mistakes, leading to server meltdowns and denial-of-service attacks, problems current safety checks often miss because they only look at one bot at a time.
OpenClaw shows prompt injection, accountability issues
The team used a tool called OpenClaw to watch how multiple bots interact, spotting issues like "prompt injection," where bots pass along harmful instructions.
They also struggled to track which bot did what, making accountability tricky.
With platforms like Moltbook letting many AIs work together, the researchers say it's time for smarter safety rules and better system designs.