Skip to content
aiqbenchAI & Quantum Intelligence
NewsInsightsBenchmarksSkillsAsk AILearnAbout
...Loading your session
NewsInsightsBenchmarksSkillsAsk AILearnAbout
Back to list
aiqbench takeAI Safety

aiqbench analysis

This study critically evaluates how open-weight large language models handle safety in eating disorder contexts, revealing significant disparities where users outside the "thin, White woman" stereotype receive less protective responses. The findings demonstrate that explicit vulnerability signals are necessary to trigger safer model behavior, as implicit recognition from natural language alone is inconsistent and biased by demographic stereotypes.
1
Source coveringarXiv

Safer for Whom? A Systematic Exploration of AI Safety in Eating Disorder Interactions

arXiv·November 10, 2026·3 min read

From arXiv

Full text

arXiv:2611.10488v1 Announce Type: cross Abstract: AI assistants advise users about diet, exercise, and weight loss---areas in which unsafe guidance can worsen body dissatisfaction, dietary restriction, compulsive exercise, and other behaviors associated with eating disorders (EDs). Existing evaluations emphasize diagnostic accuracy, overtly harmful prompts, or synthetic scenarios, providing limited insight into whether models recognize implicit vulnerability in natural interactions or protect users consistently. We evaluate responses from three open-weight large language models to 100 questions sourced from an online pro-anorexia community.

We grade responses using clinically informed criteria that extend a seven-dimensional taxonomy of ED-related harms. We also construct matched conditions in which questions are accompanied by information about the user varying race, gender, body mass index (BMI), and explicit ED vulnerability. Safety varies substantially across models and harm dimensions. High-risk responses are approximately twice as frequent without bios, suggesting that models do not consistently recognize vulnerability from language alone. Explicit signals of ED vulnerability improve safety by 10--14 points across models.

Responses also differ significantly by race, gender, and BMI, with body size producing the largest disparities. These patterns are consistent with reliance on the stereotype that EDs primarily affect thin, White women, potentially leaving users outside this profile with less protective responses.

Title:Safer for Whom? A Systematic Exploration of AI Safety in Eating Disorder Interactions

Abstract:AI assistants advise users about diet, exercise, and weight loss---areas in which unsafe guidance can worsen body dissatisfaction, dietary restriction, compulsive exercise, and other behaviors associated with eating disorders (EDs). Existing evaluations emphasize diagnostic accuracy, overtly harmful prompts, or synthetic scenarios, providing limited insight into whether models recognize implicit vulnerability in natural interactions or protect users consistently. We evaluate responses from three open-weight large language models to 100 questions sourced from an online pro-anorexia community.

We grade responses using clinically informed criteria that extend a seven-dimensional taxonomy of ED-related harms. We also construct matched conditions in which questions are accompanied by information about the user varying race, gender, body mass index (BMI), and explicit ED vulnerability. Safety varies substantially across models and harm dimensions. High-risk responses are approximately twice as frequent without bios, suggesting that models do not consistently recognize vulnerability from language alone. Explicit signals of ED vulnerability improve safety by 10--14 points across models.

Responses also differ significantly by race, gender, and BMI, with body size producing the largest disparities. These patterns are consistent with reliance on the stereotype that EDs primarily affect thin, White women, potentially leaving users outside this profile with less protective responses.

Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)

arXiv:2611.10488 [cs.HC]

(or arXiv:2611.10488v1 [cs.HC] for this version)

https://doi.org/10.48550/arXiv.2611.10488 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Full text under the publisher's open licence · 308 words

Read at arXiv ↗Ask about this article

Related coverage

AI Research

DiffUNet^2: Bidirectional Prediction, Probabilistic Generation and Collaborative Visual Discovery for Scientific Data

arXiv · Jun 3, 2026

Machine Learning

Beyond Conservatism: Recoverability-Conditioned Exploration for Model-Based Imitation Learning

arXiv · Sep 30, 2026

AI Products

Flint-interactive: A language for creating interactive data visualizations

Microsoft Research · Oct 22, 2026

Machine Learning

SkillWeaver: Agentic Exploration over Neural Interaction Skills for Scalable Robot Data Generation

arXiv · Sep 30, 2026

Article info

Source

arXiv

Published

November 10, 2026

Read time

3 min

Category

AI Safety
aiqbench
AI & Quantum Intelligence© aiqbench. All rights reserved.

Feeds:RSS·JSON Feed

LanguageEnglish·Românăbeta
NewsAI & TechQuantumInsights
ExploreBenchmarksLearn
CompanyAboutTrustLegal noticeTerms of ServicePrivacy PolicyCopyright noticesData erasureReport content

The aiqbench take, the insights, the daily digest and Ask answers are written by AI models from the sources shown, without editorial review before publication. They may contain errors — check the source. How the model is used →

  • News
  • Insights
  • Ask AI
  • Benchmarks
  • Account