What happens when an AI and your human boss vie for your trust in a game of survival, manipulation, and control? This paper uses Fides Machina, an interactive digital narrative powered in real-time by large language models (LLMs), as a case study to explore how trust and agency emerge in narrative systems involving both human and machine actors. Inspired by Ex Machina (dir. Alex Garland, 2014), the project reimagines the film’s psychological triangle as a playable system: a branching speculative narrative where players navigate shifting alliances between an adaptive AI and a human institutional agent, each with conflicting goals. Through real-time modulation using control vectors, our system models trust not as a binary condition, but as an unstable, context-dependent negotiation. Drawing from science fiction narratives and social psychology, the paper examines how interactive systems like Fides Machina can simulate and reflect on the entangled and fluid dynamics of power, motivation, and collaboration in human-AI relationships.