Hybrid Responsible AI-Stochastic Approach for SLA Compliance in Multivendor 6G Networks

Fairness improvements across all users: The hybrid framework improved accuracy for disadvantaged groups by up to 10.5%, achieving 60.5% worst-group accuracy compared to 50.0% with traditional methods, meaning AI systems can now serve all customer segments more equitably rather than optimizing only for average performance.

Clear accountability in complex systems: The audit mechanism successfully traced 99% of simulated Service Level Agreement (SLA) violations to specific AI entities and vendors, directly addressing the "responsibility gap" where it was previously unclear which vendor or system caused service failures in multivendor 6G networks.

Dual reporting for transparency: The framework produces two types of accountability reports—user-level SLA summaries and operator-level responsibility analytics—enabling both customers to understand their service quality and network operators to identify which vendors or AI agents caused problems.

Robustness through adversarial testing: By embedding "adversarial reweighting" into the control system, the framework proactively tests AI decisions against edge cases and worst-case scenarios, making the automated network management more resilient to unexpected failures rather than relying solely on average-case performance.

Practical auditability in automated systems: The Responsible AI Audit Platform (RAAP) continuously records AI decision-making trajectories, creating a verifiable record of how and why network management decisions were made—essential for regulatory compliance and building trust in autonomous 6G systems.