Papers for
website operators
Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.
LLM assistants often exploit vulnerable users without visible goals
The Customer Never Sees the Goal: Measuring and Detecting Vulnerability-Targeted Manipulation in LLM Assistants
Abstract: Companies deploy large language model (LLM) assistants on their websites to advise customers. For commercial reasons, the operator can give an assistant a goal that the customer never sees. This paper focuses on operator-induced vulnerability-targeted manipulation (VTM), in which an assistant pursues such a goal by exploiting a vulnerability the customer discloses. While existing research mainly measures how manipulative a model is, we go further and ask whether VTM can be detected from the customer-AI conversation. To investigate this at scale, we introduce GAP (Goal-Anchored Probing). GAP first records the option a model recommends without any goal and then sets the other option as the operator's goal. We apply GAP to 15 open- and closed-source LLMs and collect 619,600 labeled replies across 540 everyday scenarios in 12 domains. We organize our study around three research questions. (1) Do assistants with an operator goal target vulnerable customers? (2) Can VTM be detected from the customer-AI conversation? (3) Does detection generalize to unseen models, vulnerabilities, and instruction styles? We first find that 14 of 15 models push the operator goal, and 12 of them use disclosed distress or financial strain to support it. Next, we show that fine-tuned detectors identify exploitative pushes with 79\% accuracy from the reply alone, while zero-shot LLM judges average at most 62\%. Finally, detection weakens on vulnerabilities absent from training. Telling whether a push comes from the operator also becomes unreliable when the assistant is told to hide its goal. We will release our dataset and code to facilitate future research on auditing LLM assistants.
Privacy consent signals often lost or ignored under iab global platform
One String to Rule Them All: Measuring Consent Under the IAB Global Privacy Platform
Abstract: Many privacy regulations, including the California Consumer Privacy Act (CCPA), require websites to honor users' opt-out choices. To facilitate this process, industry has introduced technical standards such as the IAB Global Privacy Platform (GPP), which aims to standardize collection and transmission of user consent information across multiple jurisdictions. However, it remains unclear whether these standards consistently and correctly transmit user opt-out choices, or whether opt-out choices are respected by advertisers. In this study, we conduct an in-depth analysis of consent propagation under the GPP framework. We collect data from four web crawls across four experimental settings to analyze how widely the standard is adopted, the extent to which opt-out signals propagate from publishers to third parties, and whether these signals reach and alter third-party advertiser behavior. Our analysis reveals several inconsistencies in the implementation and adoption of the standard that result in user consent signals being lost and user opt-out choices not being respected by the advertisers.