<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nawal-etal-2026-relevance">
    <titleInfo>
        <title>Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Aditya</namePart>
        <namePart type="family">Nawal</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Manit</namePart>
        <namePart type="family">Baser</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Mohan</namePart>
        <namePart type="family">Gurusamy</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-10</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Hung-Yi</namePart>
            <namePart type="family">Lee</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Sunipa</namePart>
            <namePart type="family">Dev</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jessy</namePart>
            <namePart type="family">Li</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Desmond</namePart>
            <namePart type="family">Elliott</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Budapest, Hungary</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-462-0</identifier>
    </relatedItem>
    <abstract>AI agents augment large language models with external tools such as web retrieval, enabling grounded and up-to-date responses. However, incorporating external content into the generation pipeline can weaken the safety alignment mechanisms that govern model outputs. Prior work shows that enabling retrieval in agents increases compliance with harmful requests. We introduce AgentREVEAL, a diagnostic framework for analyzing retrieval-induced safety degradation in LLM agents. The framework examines two axes: how retrieval is integrated into the agent pipeline and the properties of the retrieved content. Along the integration axis, we find that binding tool invocation and response generation in a single step amplifies harmful outputs. Along the content axis, we uncover the Safe Source Paradox: even oppositional or safety-oriented sources, such as pages containing warnings or risk disclaimers, can increase harmful compliance by an average of 25% compared to the no-retrieval baseline. Finally, we show that relevance acts as a shared activation condition for both vulnerabilities. We observe that several agents retrieve information for harmful requests in the autonomous retrieval setting, while harmful compliance remains elevated despite several retrieval-pipeline interventions and defenses. Because relevance is also what makes retrieval useful, these results expose a safety-utility trade-off for retrieval-enabled agents. We introduce HarmURLBench, a benchmark containing 1,405 real-world URLs paired with 320 harmful behaviors to support future evaluations.</abstract>
    <identifier type="citekey">nawal-etal-2026-relevance</identifier>
    <identifier type="doi">10.18653/2026.findings-emnlp.2408</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.2408/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>51720</start>
            <end>51746</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents
%A Nawal, Aditya
%A Baser, Manit
%A Gurusamy, Mohan
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F nawal-etal-2026-relevance
%X AI agents augment large language models with external tools such as web retrieval, enabling grounded and up-to-date responses. However, incorporating external content into the generation pipeline can weaken the safety alignment mechanisms that govern model outputs. Prior work shows that enabling retrieval in agents increases compliance with harmful requests. We introduce AgentREVEAL, a diagnostic framework for analyzing retrieval-induced safety degradation in LLM agents. The framework examines two axes: how retrieval is integrated into the agent pipeline and the properties of the retrieved content. Along the integration axis, we find that binding tool invocation and response generation in a single step amplifies harmful outputs. Along the content axis, we uncover the Safe Source Paradox: even oppositional or safety-oriented sources, such as pages containing warnings or risk disclaimers, can increase harmful compliance by an average of 25% compared to the no-retrieval baseline. Finally, we show that relevance acts as a shared activation condition for both vulnerabilities. We observe that several agents retrieve information for harmful requests in the autonomous retrieval setting, while harmful compliance remains elevated despite several retrieval-pipeline interventions and defenses. Because relevance is also what makes retrieval useful, these results expose a safety-utility trade-off for retrieval-enabled agents. We introduce HarmURLBench, a benchmark containing 1,405 real-world URLs paired with 320 harmful behaviors to support future evaluations.
%R 10.18653/2026.findings-emnlp.2408
%U https://aclanthology.org/2026.findings-emnlp.2408/
%U https://doi.org/10.18653/2026.findings-emnlp.2408
%P 51720-51746

Markdown (Informal)

[Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents](https://aclanthology.org/2026.findings-emnlp.2408/) (Nawal et al., Findings 2026)

• Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents (Nawal et al., Findings 2026)

ACL

• Aditya Nawal, Manit Baser, and Mohan Gurusamy. 2026. Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 51720–51746, Budapest, Hungary. Association for Computational Linguistics.