@inproceedings{le-jeune-etal-2025-realharm,
    title = "{R}eal{H}arm: A Collection of Real-World Language Model Application Failures",
    author = "Le Jeune, Pierre  and
      Liu, Jiaen  and
      Rossi, Luca  and
      Dora, Matteo",
    editor = "Derczynski, Leon  and
      Novikova, Jekaterina  and
      Chen, Muhao",
    booktitle = "Proceedings of the The First Workshop on LLM Security (LLMSEC)",
    month = aug,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-emnlp/2025.llmsec-1.7/",
    pages = "87--100",
    ISBN = "979-8-89176-279-4",
    abstract = "Language model deployments in consumer-facing applications introduce numerous risks. While existing research on harms and hazards of such applications follows top-down approaches derived from regulatory frameworks and theoretical analyses, empirical evidence of real-world failure modes remains underexplored. In this work, we introduce RealHarm, a dataset of annotated problematic interactions with AI agents built from a systematic review of publicly reported incidents. Analyzing harms, causes, and hazards specifically from the deployer{'}s perspective, we find that reputational damage constitutes the predominant organizational harm, while misinformation emerges as the most common hazard category. We empirically evaluate state-of-the-art guardrails and content moderation systems to probe whether such systems would have prevented the incidents, revealing a significant gap in the protection of AI applications."
}