The growing autonomy and self-learning abilities of advanced technologies alter the dynamics of human-AI interaction, while also raising important ethical concerns. This study calls for rethinking AI literacy to encompass additional abilities necessary for responsible human-AI interaction. Beyond enhancing efficiency and performance, research on AI literacy must address unintended consequences and risks, and thus requires a digital responsibility perspective. Through interviews with different groups of knowledge workers, the study outlines factors that enable and influence knowledge workers to interact responsibly with AI technologies. Our research uncovers tensions concerning the concept of AI literacy that may challenge responsible human-AI interaction. These concern especially problems related to skill development and responsibility attribution. These problems emphasize the need for shared responsibilities and to ensure alignment between knowledge workers’ abilities and their roles – what we call the “role-ability fit.” We further offer recommendations to promote responsible human-AI interaction.