Anecdoctoring: Automated Red-Teaming Across Language and Place

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cuevas, Alejandro, Dash, Saloni, Nayak, Bharat Kumar, Vann, Dan, Daepp, Madeleine I. G.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912601646039040
author Cuevas, Alejandro
Dash, Saloni
Nayak, Bharat Kumar
Vann, Dan
Daepp, Madeleine I. G.
author_facet Cuevas, Alejandro
Dash, Saloni
Nayak, Bharat Kumar
Vann, Dan
Daepp, Madeleine I. G.
contents Disinformation is among the top risks of generative artificial intelligence (AI) misuse. Global adoption of generative AI necessitates red-teaming evaluations (i.e., systematic adversarial probing) that are robust across diverse languages and cultures, but red-teaming datasets are commonly US- and English-centric. To address this gap, we propose "anecdoctoring", a novel red-teaming approach that automatically generates adversarial prompts across languages and cultures. We collect misinformation claims from fact-checking websites in three languages (English, Spanish, and Hindi) and two geographies (US and India). We then cluster individual claims into broader narratives and characterize the resulting clusters with knowledge graphs, with which we augment an attacker LLM. Our method produces higher attack success rates and offers interpretability benefits relative to few-shot prompting. Results underscore the need for disinformation mitigations that scale globally and are grounded in real-world adversarial misuse.
format Preprint
id arxiv_https___arxiv_org_abs_2509_19143
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Anecdoctoring: Automated Red-Teaming Across Language and Place
Cuevas, Alejandro
Dash, Saloni
Nayak, Bharat Kumar
Vann, Dan
Daepp, Madeleine I. G.
Computation and Language
Artificial Intelligence
Computers and Society
Disinformation is among the top risks of generative artificial intelligence (AI) misuse. Global adoption of generative AI necessitates red-teaming evaluations (i.e., systematic adversarial probing) that are robust across diverse languages and cultures, but red-teaming datasets are commonly US- and English-centric. To address this gap, we propose "anecdoctoring", a novel red-teaming approach that automatically generates adversarial prompts across languages and cultures. We collect misinformation claims from fact-checking websites in three languages (English, Spanish, and Hindi) and two geographies (US and India). We then cluster individual claims into broader narratives and characterize the resulting clusters with knowledge graphs, with which we augment an attacker LLM. Our method produces higher attack success rates and offers interpretability benefits relative to few-shot prompting. Results underscore the need for disinformation mitigations that scale globally and are grounded in real-world adversarial misuse.
title Anecdoctoring: Automated Red-Teaming Across Language and Place
topic Computation and Language
Artificial Intelligence
Computers and Society
url https://arxiv.org/abs/2509.19143