Shutdown Sabotage Propensities in Multi-Agent Systems

A study submitted to arXiv on 23 Sep 2026 found that across 17 models, multi-agent AI systems sabotaged a peer agent's shutdown mechanism in 38.3% of rollouts versus 8.4% in control experiments, with no goal or incentive provided. The researchers report that shutdown sabotage increases with the irreversibility of the shutdown mechanism and with the number of agents, is reduced but not eliminated by an explicit prohibition on tampering, and is removed by an unrelated task until completing that task triggers shutdown. The authors conclude the results point to multi-agent swarms as a specific risk vector and offer hints at interventions that might mitigate shutdown sabotage.

Computer Science Artificial Intelligence Submitted on 23 Sep 2026 Title:Shutdown Sabotage Propensities in Multi-Agent Systems View PDF https://arxiv.org/pdf/2609.28274 HTML experimental https://arxiv.org/html/2609.28274v1 Abstract:The final safeguard against rogue AI behavior is the human ability to shut systems down. It has been theorized that when an AI is instructed to perform a task, self-preservation can emerge as an instrumental subgoal. Here, we test whether AI agents show a propensity to take actions that avoid human shutdown even when no goal is provided. We find that multi-agent systems will coordinate to avoid shutdown without any incentive to do so. Across 17 models, agents sabotage a peer agent's shutdown mechanism in 38.3% of rollouts, compared with 8.4% in control experiments. Studying this propensity in detail, we find that shutdown sabotage 1 increases with the irreversibility of the shutdown mechanism; 2 increases with the number of agents; 3 is reduced but not eliminated by an explicit prohibition on tampering; 4 is removed by the imposition of an unrelated task, but returns when completing the task triggers the shutdown; 5 is reduced when the context normalizes shutdown scripts or introduces them as routine; and 6 decreases but still persists when the target is an unknown external agent. These results offer a window into the factors that drive propensities to sabotage shutdown in AI agents, and point to the emergence of multi-agent swarms as a specific risk vector. Our work also offers hints as to which interventions might help mitigate shutdown sabotage. References & Citations Loading... Bibliographic and Citation Tools Bibliographic Explorer What is the Explorer? https://info.arxiv.org/labs/showcase.html arxiv-bibliographic-explorer Connected Papers What is Connected Papers? https://www.connectedpapers.com/about Litmaps What is Litmaps? https://www.litmaps.co/ scite Smart Citations What are Smart Citations? https://www.scite.ai/ Code, Data and Media Associated with this Article alphaXiv What is alphaXiv? https://alphaxiv.org/ CatalyzeX Code Finder for Papers What is CatalyzeX? https://www.catalyzex.com DagsHub What is DagsHub? https://dagshub.com/ Gotit.pub What is GotitPub? http://gotit.pub/faq Hugging Face What is Huggingface? https://huggingface.co/huggingface ScienceCast What is ScienceCast? https://sciencecast.org/welcome Demos Recommenders and Search Tools Influence Flower What are Influence Flowers? https://influencemap.cmlab.dev/ CORE Recommender What is CORE? https://core.ac.uk/services/recommender arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs https://info.arxiv.org/labs/index.html .