Researchers at NAIL Institute release AVE Database cataloging 50 distinct failure modes to improve safety in multi-agent AI systems

AVE Database provides an open taxonomy of 50 failure modes specific to multi-agent AI systems

Resource aims to help developers identify and mitigate risks in AI systems with multiple interacting agents

Published by NAIL Institute as a reference tool for AI safety and system design

Open taxonomy approach allows community contributions to expand failure mode documentation