Naive perspectives on AI capabilities can diminish students’ epistemic curiosity for exploring AI ethics. While foundational AI knowledge can theoretically enable meaningful knowledge gap recognition, an essential prerequisite for curiosity activation, its effects in AI ethics instruction remain unexplored. This study examines whether foundational AI knowledge affects epistemic curiosity and self-efficacy developmental patterns during game-based AI ethics instruction. A quasi-experimental design compared 148 students (M age = 13.5 years) receiving either AI fundamentals video pre-training followed by game-based AI ethics learning or game-based learning alone. 2x3 mixed repeated measures ANCOVA was used to examine developmental trajectories of in-game epistemic curiosity and self-efficacy measurements while controlling for prior AI ethics knowledge. Results show that students who received pre-training demonstrated significantly different epistemic curiosity patterns, showing growth compared to the game-only condition (d = 0.34 at final measurement). Self-efficacy development remained equivalent between conditions. These findings contribute to AI literacy framework development by providing empirical justification for integrated approaches linking AI fundamentals with AI ethics instruction to support epistemic curiosity sustainability. Additionally, this research offers empirical insights into the design of game-based learning components for AI literacy education.