HelpBench: Assessing the Ability of LLMs to Provide Privacy, Safety, and Security Advice
2026-06-23 • Cryptography and Security
Cryptography and Security
AI summaryⓘ
The authors created HelpBench, a set of 450 real-life questions to test how well large language models (LLMs) can give correct advice on digital privacy, safety, and security. They made guidelines to check if the answers were accurate and polite. Testing 18 top LLMs showed that the models usually gave good advice but sometimes gave wrong or harmful information about 10% of the time. The authors highlight the importance of fixing these errors for LLMs to be trusted helpers in privacy and security matters.
Large Language ModelsDigital PrivacyCybersecurityTwo-Factor AuthenticationPhishingAuto-RaterAccount RecoverySecurity BenchmarksFactual Accuracy
Authors
Sarah Meiklejohn, Sunny Consolvo, Patrick Gage Kelley, Tara Matthews, Sai Teja Peddinti, Renee Shelby, Lenin Simicich, Kurt Thomas
Abstract
This paper introduces HelpBench, a benchmark for assessing whether LLMs are capable of providing accurate help in response to questions about digital privacy, safety, and security. We curated 450 questions representing authentic user situations and developed rubrics for each question to evaluate the factual accuracy and tone of a response. Example questions touch on how to regain access to lost or suspended accounts, how to balance the trade-offs of hardware security keys versus other forms of two-factor authentication, whether a suspicious email is likely a scam, or whether an abuser might be able to track an individual based on their device peripherals. We then developed and applied an auto-rater to evaluate responses from 18 state-of-the-art LLMs. Our results indicate that while models provide high-quality advice (with scores of 82% on average), one in ten responses from models scores less than 65%, reflecting inaccurate and even harmful advice. Addressing these failures is critical for models to serve as trustworthy sources of assistance for digital privacy, safety, and security needs.