Closing the Responsibility Gap in AI-based Network Management: An Intelligent Audit System Approach

AI network management systems improve performance but create accountability gaps: While AI-based tools reduce network downtime and operating costs, they introduce new risks including privacy violations, algorithmic bias, and model inaccuracies—yet there's often unclear responsibility when problems occur.

The paper proposes an "intelligent audit system" to assign responsibility to AI agents: Rather than blaming the entire network when something goes wrong, the framework identifies which specific AI management agent caused the problem and quantifies their level of responsibility, enabling more targeted accountability.

The audit system achieved strong accuracy rates in testing: The deep learning component identified problematic AI agents with 96% accuracy, while the machine learning component tracked network conditions with 83% accuracy, suggesting the approach is technically viable.

Human oversight gaps are a key concern: The paper highlights that removing human supervision from network management—one of AI's main benefits for cost reduction—simultaneously removes an important safeguard, making automated accountability mechanisms increasingly necessary.

This framework addresses a practical governance challenge: By creating a system that can pinpoint which AI agent is responsible for network failures or issues, organizations can better enforce accountability and potentially hold specific AI systems (rather than operators or companies broadly) responsible for their decisions.