As generative AI becomes increasingly integrated into modern applications, traditional identity security models are being challenged, particularly in how permissions are managed for dynamic access control initiated by AI identities. Traditional methods focused on ingress traffic, but the unpredictability of AI agents necessitates a shift to also managing egress traffic. Tools like Lunar.Dev serve as an open-source API gateway for outbound requests, enabling real-time policy enforcement through integration with authorization services like Permit.io, Open Policy Agent, and OPAL. This approach allows for proactive access control, where AI agents can dynamically request additional permissions as needed. The shift from Mandatory and Discretionary Access Control to Proactive Access Control is essential to accommodate the evolving needs of AI systems without compromising security. This proactive, dynamic strategy streamlines permissions and decentralizes enforcement, ensuring consistency and flexibility in managing AI identities. The article is part of a series addressing the complexities of AI in Identity and Access Management, with future installments set to explore further aspects of AI identity permissions.