The collaboration marks a transition from static model scanning to dynamic runtime defense. While HiddenLayer previously focused on assessing model artifacts through Unity Catalog before deployment, the new integration embeds security deeper into the operational path. This allows organizations to monitor live interactions between models, agents, tools, and Model Context Protocol servers in real time.

Chris Sestito, CEO and co-founder of HiddenLayer, noted that many companies are deploying autonomous systems without the necessary safeguards to manage evolving risks. The platform addresses this by layering AI-native threat detection—such as prompt injection, model theft, and data leakage—directly into existing Databricks workflows. Stephen Orban, SVP of Product Partnerships at Databricks, emphasized that this move extends governance beyond simple access controls, providing the visibility required to enforce guardrails across increasingly interconnected AI environments.