A new study demonstrates that using diverse ensembles of AI monitors significantly improves the detection of misaligned actions in autonomous systems, outperforming traditional methods that rely on scale. This approach emphasizes the importance of monitor diversity and fine-tuning for enhancing safety in AI applications at manageable costs.