Blocking Slow-Burn Attacks with Contextual Policies in Omnigent

Databricks introduces contextual policies that evaluate sequences of agent actions rather than single actions to detect slow-burn attacks. The approach aims to improve security for agentic AI systems by considering the broader context of behavior over time.