Azure AI Foundry integrates several security features to help organizations manage risks effectively. It includes Spotlighting to detect and block prompt injection attacks, real-time task adherence evaluation to ensure agents operate within defined boundaries, and continuous evaluation and monitoring of agentic systems to track performance and safety. Additionally, Microsoft Defender for Cloud provides integrated security alerts and recommendations directly within the Azure AI Foundry portal, allowing developers to identify and address threats in real time.

Microsoft Purview enhances data security for AI applications by providing a comprehensive set of tools for data security posture management. It allows developers to integrate data security and compliance controls seamlessly into their AI applications. With the Purview SDK, developers can automatically identify and block sensitive data before it is exposed, ensuring compliance with regulatory requirements. Furthermore, Purview's integration with Azure AI Foundry enables security teams to gain visibility into data risks and manage compliance effectively without additional development work.

Task adherence is crucial for AI agents to ensure they operate within defined instructions and policy boundaries. Deviations can lead to misuse of tools or unintended exposure of sensitive data. Azure AI Foundry addresses this by introducing a real-time task adherence evaluation API that assesses an agent's behavior against its assigned tasks. If an agent goes off-task, the system can block tool use or trigger a human review, helping organizations maintain control and compliance throughout the AI development lifecycle.