Newest papers

Data-Void Exploits in Web-Searching LLMs

Michele Guerra, Wuyeh Jobe, Sanjana Nambiar, Christina Pöpper, Azza Abouzied · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.30365 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Em vez de navegar através de listas de links classificadas, os usuários lêem respostas sintetizadas. Esta mudança cria uma nova vulnerabilidade do LLM: uma exploração de dados em vácuo. Os desinformadores publicam suas narrativas em espaços de pesquisa clássicos, onde os usuários vêem dicas de classificação e resultados concorrentes promissores, os LLMs podem obscurecer quando uma resposta se baseia em evidências superficiais escassas, não verificadas ou fontes adversárias. Nós estudamos esta ameaça aumentando o tom da escolha do usuário na web - cético, excitado ou explorado - ou listando essas páginas em suas respostas a promessas de busca de dados em vácuo. Ao contrário da pesquisa clássica, onde os usuários veem dicas de classificação e resultados promissores, os LLMs podem obs

Machine translation (Meta M2M100); the English original follows.

Abstract

Web-searching Large Language Models (LLMs) have changed how users seek online information: instead of sifting through ranked lists of links, users read synthesized answers. This shift creates a new LLM vulnerability: a data-void exploit. Disinformers publish their narratives in search spaces where relevant and reliable content is scarce, i.e. data voids, then rely on the LLM's web-search to retrieve and then use, cite, or list those pages in its responses to data-void seeking prompts. Unlike classical search, where users see ranking cues and competing results, LLMs can obscure when an answer relies on sparse, unverified, or adversarial sources. We study this threat by varying the user's prompt tone - skeptical, excited, or neutral - and LLM choice - ChatGPT, Gemini, or Sonar Pro - as well as three attacker controls. The attacker can use one of five known strategies to construct and fill a data void with their narrative, host their content at varying levels of visibility from well-established, indexed domains to less established ones, and include generative-optimization engine features thought to boost citing by web-searching LLMs. We find that data-void exploits can succeed with only a single attacker controlled page. Across all conditions, attacker-content appears in LLM responses 31% of the time. User controls offer limited protection: while skeptical prompts slightly reduce success and slightly increase flagging or warnings in LLM responses, we find no evidence of large effects of prompt tone or model choice on success rates, although smaller effects cannot be ruled out. Overall, our results establish data voids as a practical inference-time attack surface for manipulating web-searching LLMs through strategically placed web content and highlight the urgent need for defenses against such exploits.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Exploit, Computer science, Flagging, Computer security, Visibility, Internet privacy

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange