An organization deploys an AI recruitment platform to screen job applicants. The IS auditor identifies that the platform's decisions may be influenced by model bias. Which of the following risk…

Implementing a periodic bias testing process with parameter adjustments addresses model bias as a continuous risk management activity, which is appropriate for a production system where bias may evolve as data changes. This approach follows a risk-based control framework by establishing monitoring and correction mechanisms without unnecessarily halting operations. It is proportionate, sustainable, and aligns with responsible AI governance practices.

Suspending the system is a disproportionate response that disrupts operations and does not constitute a mitigation strategy; it is a temporary halt, not a control.

Retraining on a single externally certified dataset does not address ongoing bias risk and provides no mechanism for detecting or correcting future drift or emerging bias.

Requiring manual review of all AI decisions eliminates the efficiency benefit of the AI system and is not a scalable or practical long-term mitigation strategy.