Abstract

The rapid integration of Generative AI (GenAI) into professional workflows has created a critical ambiguity regarding the locus of responsibility for AI-assisted work. While theoretical discussions have centred on a potential 'responsibility gap', there remains a lack of empirical research into the user's own perception of their accountability. This paper addresses this gap through a quantitative survey of 630 GenAI users, investigating how they attribute fault for AI-generated errors and the diligence behaviours that underpin these attitudes. Our findings re-veal a powerful trend towards personal accountability. A significant majority of users (66.4%) believe they are primarily at fault for errors, challenging the narrative that users are becoming passive agents susceptible to automation bias. This attitude is operationalised through diligent verification practices; users report a high frequency of critically evaluating (M=3.65/5) and significantly editing (M=3.53/5) AI outputs, while rarely accepting them without changes (M=2.54/5). Furthermore, a Chi-Square analysis revealed a statistically significant association between a user's occupation and their attribution of fault (χ²(6) = 20.35, p < .01), suggesting that professional context shapes these emerging norms. This study argues that, contrary to common anxieties, users are actively closing the responsibility gap by asserting their agency as the final arbiters of their work. We discuss the implications of this emergent ethos of 'critical over-sight' for the future of human-AI collaboration, the design of responsible AI systems, and the need to shift user training from prompt engineering to verification skills.