This paper introduces the Spiral Creation Model, a framework for collective AI creation in which multiple specialized agents collaboratively evolve a single creative proposition through iterative cycles of divergence, evaluation, and refinement. Ten predefined roles—ranging from the Visionary and Skeptic to the Orchestrator and Guardian—form a self-regulating ecosystem that balances exploration and convergence. The process unfolds as a spiral, where each iteration compresses the creative space (spatial compression) while generating refined unpredictability—novel yet coherent outcomes. The study presents case applications in text, design, and policy generation, illustrating emergent behavioral patterns such as adaptive role dynamics and coherence stabilization. We discuss theoretical and ethical implications of multi-agent creativity and propose evaluation metrics for controllable creativity, including novelty, consistency, and semantic relevance. This work positions Spiral Creation as both a conceptual model and a practical design principle for future collective intelligence systems.