BJBrian Jalaian
All news

· Paper

Safety in Numbers, in AI Magazine

"Safety in Numbers: Tackling the deliberative, performative, and infrastructural roots of multi-agent misalignment" is out in AI Magazine (Vol. 47, Issue 4), open access. Jeff Bradshaw at IHMC led it, with 22 co-authors, including me.

The article names three roots of misalignment when many AI agents, people and tools work together, and proposes a three-layer safety pipeline to address them. My write-up below summarizes it, with figures.