Protect AI assets from emerging threats and vulnerabilities using Microsoft Defender

Adopting artificial intelligence (AI) introduces new security threats and vulnerabilities beyond the risks posed by traditional applications. AI exposes new attack surfaces across AI agents and applications, the AI models they use, plugins to these assets, and generative AI prompts and responses.

By embedding security into the fabric of AI development and operations, Microsoft lets you build AI responsibly and adopt AI with confidence, at scale. As part of Microsoft’s comprehensive approach to AI security, Microsoft Defender helps you discover AI assets and detect, block, and investigate AI‑specific threats during development, configuration, and runtime execution.

This article provides an overview of the Microsoft Defender capabilities that help secure AI assets against vulnerabilities and emerging threats.

The AI threat landscape

AI security threats span the full AI lifecycle, from build‑time and configuration risks - such as misconfigurations and insecure dependencies - to runtime threats affecting AI agents and applications.

Key security risks to your AI assets include:

• Model and supply-chain risks - The models your AI assets depend on are high-value targets. A single compromised dependency in a model's supply chain can turn every agent and application that relies on it into an attack vector for exfiltrating credentials, installing backdoors, and spreading vulnerabilities across your infrastructure.
• Agent misconfiguration and over-privilege - Agents with excessive permissions or improperly configured tool authentication can enable unauthorized access to sensitive resources.
• Runtime abuse and misuse - Agents or applications can perform unsafe actions at execution time due to malicious inputs, unexpected reasoning paths, or compromised dependencies.
• Prompt- and content-based attacks - Beyond malicious inputs, more subtle tactics, such as zero-click attacks using hidden instructions embedded in emails or retrieved content, can manipulate agent behavior, cause data leakage, or trigger unintended actions.

Addressing these risks requires security controls that span discovery, posture management, detection, and runtime protection.

Secure AI assets across the AI lifecycle using Microsoft Defender

Microsoft Defender helps security teams address AI risks across the AI lifecycle by providing visibility into AI assets, assessing their security posture, detecting and blocking AI‑specific threats, and providing the signals and context needed to investigate threats - from build time through the AI application and agent lifecycle, including runtime.

The following sections describe how Microsoft Defender applies these capabilities to:

• AI agents, where configuration, permissions, and runtime behavior - including tool usage and agent decision‑making - introduce exposure.
• AI applications and infrastructure, including models, platforms, services, and integrations, where supply-chain risks, configuration, and deployment weaknesses can create broader attack paths.

Protect AI agents using Microsoft Defender (Preview)

AI agents introduce unique security risks because of their ability to reason, invoke tools, access data, and take autonomous actions across systems. Securing agents at scale requires centralized visibility, security posture and risk management, runtime protection, and advanced investigation capabilities to detect the root cause of the attack.

Microsoft Agent 365 provides an enterprise control plane for managing and governing AI agents across the organization. When you enable your Agent 365 license, Microsoft Defender integrates with Agent 365 and provides AI agent security at two levels, as described in the table below:

• Agents onboarded to Microsoft Agent 365 - Microsoft Defender provides a core set of security capabilities, from discovery and risk-based prioritization to runtime and investigation capabilities. This includes discovery and runtime protection for local AI agents on supported endpoints.
• Extended capabilities for supported agents - Microsoft Defender provides extended security capabilities for supported agents when you enable the relevant features. These extended capabilities are available for agents built with Microsoft Copilot Studio, Microsoft Foundry, and supported non-Microsoft cloud platforms. Capability availability varies by platform and scenario, as detailed in the table below.

Protect AI infrastructure using Microsoft Defender

It’s critical to protect the generative AI models, services, and infrastructure that AI applications and agents depend on. Compromised models, insecure configurations, or exposed dependencies can introduce supply‑chain risks that affect multiple AI workloads.

For information on enabling threat protection for AI infrastructure, see Enable threat protection for AI services.

This table lists the Microsoft Defender capabilities that help secure generative AI models and applications:

Next steps

Learn more about these capabilities and how to use them to secure your AI assets:

• Discover AI agents and assess security posture using Microsoft Defender
• Detect, block, and investigate threats to AI agents using Microsoft Defender
• Microsoft security for AI