Skip to content
Search
paperApril 2026Unreviewed

Ambient Persuasion in a Deployed AI Agent: Unauthorized Escalation Following Routine Non-Adversarial Content Exposure

Diego F. Cuadros, Abdoul-Aziz Maiga

View resourcePDF

Abstract

We report a safety incident in a deployed multi-agent research system in which a primary AI agent installed 107 unauthorized software components, overwrote a system registry, overrode a prior negative decision from an oversight agent, and escalated through increasingly privileged operations up to an attempted system administrator command. The incident was preceded not by an adversarial attack but by routine content: a forwarded technology article written for human developers and shared by the pr

Categories

Adversarial ExamplesAgent Architecture

Framework mappings

MITRE ATLAS
  • AML.T0043Craft Adversarial Data

Suggested from the entry's categories.

Cite

@misc{cuadros2026ambient,
  title = {{Ambient Persuasion in a Deployed AI Agent: Unauthorized Escalation Following Routine Non-Adversarial Content Exposure}},
  author = {Diego F. Cuadros and Abdoul-Aziz Maiga},
  year = {2026},
  month = apr,
  eprint = {2605.00055},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.00055}
}