[2603.26676] Evaluating Human-AI Safety: A Framework for Measuring Harmful Capability Uplift

Computer Science > Computers and Society arXiv:2603.26676 (cs) [Submitted on 6 Mar 2026] Title:Evaluating Human-AI Safety: A Framework for Measuring Harmful Capability Uplift Authors:Michelle Vaccaro, Jaeyoon Song, Abdullah Almaatouq, Michiel A. Bakker View a PDF of the paper titled Evaluating Human-AI Safety: A Framework for Measuring Harmful Capability Uplift, by Michelle Vaccaro and 3 other authors View PDF HTML (experimental) Abstract:Current frontier AI safety evaluations emphasize static benchmarks, third-party annotations, and red-teaming. In this position paper, we argue that AI safety research should focus on human-centered evaluations that measure harmful capability uplift: the marginal increase in a user's ability to cause harm with a frontier model beyond what conventional tools already enable. We frame harmful capability uplift as a core AI safety metric, ground it in prior social science research, and provide concrete methodological guidance for systematic measurement. We conclude with actionable steps for developers, researchers, funders, and regulators to make harmful capability uplift evaluation a standard practice. Subjects: Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC) Cite as: arXiv:2603.26676 [cs.CY]   (or arXiv:2603.26676v1 [cs.CY] for this version)   https://doi.org/10.48550/arXiv.2603.26676 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Michelle Vaccaro [view email] [v1...