Safety in numbers: Tackling the deliberative, performative, and infrastructural roots of multi‐agent misalignment
AI Magazine · 2026-12 · 47 (4)
DOI: 10.1002/aaai.70124 ↗Kurz gefasst
Fachartikel in der Zeitschrift AI Magazine, online erschienen am 08.11.2026. Verfasst von Jeffrey M. Bradshaw und 22 weiteren Autor:innen. Die Arbeit nennt 116 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Multi‐agent systems (MAS)—a uniquely important subset of the wider field commonly called “Agentic AI”—enable complex work processes involving distributed networks of multiple AIs, people, and technology tools. However, they also carry big risks. Root causes of catastrophic safety failures in multi‐agent collaboration (as in human collaboration) can be classed into three overlapping categories: 1. Deliberative —deviating from shared MAS commitments to fulfill explicitly expressed joint intentions; 2. Performative —inability to coordinate well or to adapt to changing circumstances; and 3. Infrastructural —safety, reliability, and trust in the operating environment. In the spirit of the international Multi‐Agent System Safety and Teamwork (MASST) initiative (Bradshaw et al., 2026), our multi‐disciplinary research team seeks to tackle the root causes of these three challenges through development, experimentation, and deployment of an integrated multi‐layered safety pipeline, enabling “safety in numbers” as increasingly powerful multi‐agent systems are deployed. Although this proposal is merely an initial sketch, it is our hope that sharing these ideas will inspire others to contribute to the discussion as we continue moving forward. We look forward not only to exciting discoveries within laboratory settings while also being committed to improving safe rollout of large‐scale MAS into the world to perform work in high‐consequence applications.
Zitieren
APA
Bradshaw, J. M., Johnson, M., Suri, N., Amato, A., Venable, K. B., Montanari, R., Romandini, N., Fronteddu, R., Bunch, L., Lott, J., Beaumont, W. D., Uszok, A., Cavicchi, S., Vignati, M., Shinnick, B., Jalaian, B., Pinelis, J., Goodrich, M. A., Diggelen, J. V., …, & Fernandez, M. T. J. (2026). Safety in numbers: Tackling the deliberative, performative, and infrastructural roots of multi‐agent misalignment. AI Magazine, 47(4). https://doi.org/10.1002/aaai.70124
Vancouver
Bradshaw JM, Johnson M, Suri N, Amato A, Venable KB, Montanari R, et al. Safety in numbers: Tackling the deliberative, performative, and infrastructural roots of multi‐agent misalignment. AI Magazine. 2026;47(4). doi:10.1002/aaai.70124
BibTeX
@article{Bradshaw2026,
author = {Bradshaw, Jeffrey M. and Johnson, Matthew and Suri, Niranjan and Amato, Alessandro and Venable, K. Brent and Montanari, Rebecca and Romandini, Nicolò and Fronteddu, Roberto and Bunch, Larry and Lott, James and Beaumont, William de and Uszok, Andrzej and Cavicchi, Sara and Vignati, Mica and Shinnick, Ben and Jalaian, Brian and Pinelis, Jane and Goodrich, Michael A. and Diggelen, Jurriaan van and Woods, David D. and Hayes, Patrick J. and Warren, Margaret and Fernandez, Manuele Tele Junior},
title = {Safety in numbers: Tackling the deliberative, performative, and infrastructural roots of multi‐agent misalignment},
journal = {AI Magazine},
year = {2026},
volume = {47},
number = {4},
doi = {10.1002/aaai.70124}
}
EUVIMED