Microsoft Security Blog: Securing DeepSeek and other AI systems with Microsoft Security

Feb 13, 2025

—

Source URL: https://www.microsoft.com/en-us/security/blog/2025/02/13/securing-deepseek-and-other-ai-systems-with-microsoft-security/
Source: Microsoft Security Blog
Title: Securing DeepSeek and other AI systems with Microsoft Security

Feedly Summary: Microsoft Security provides cyberthreat protection, posture management, data security, compliance and governance, and AI safety, to secure AI applications that you build and use. These capabilities can also be used to secure and govern AI apps built with the DeepSeek R1 model and the use of the DeepSeek app.
The post Securing DeepSeek and other AI systems with Microsoft Security appeared first on Microsoft Security Blog.

AI Summary and Description: Yes

**Summary:** The text highlights the critical importance of security in the development and deployment of AI applications. It emphasizes Microsoft Security’s offerings, such as threat protection, compliance, and governance, specifically for AI models like DeepSeek R1. It discusses various security measures, including posture management, content safety, and data loss prevention, designed to safeguard AI workloads against cyber threats.

**Detailed Description:**
The text outlines a comprehensive strategy for securing AI applications, particularly focusing on Microsoft’s solutions and methodologies:

– **Security Foundation for AI Applications**: The piece begins by stressing that a robust security foundation is essential for successful AI transformation. With the exponential growth of AI usage, visibility into AI applications and tools becomes crucial.

– **DeepSeek R1 Model**:
– DeepSeek R1 is integrated into Azure AI Foundry and GitHub, offering over 1,800 models.
– The model has undergone rigorous evaluations to mitigate potential security risks.
– Microsoft ensures data protection within Azure’s secure environment.

– **Azure AI Content Safety**:
– Built-in filtering tools help detect and block harmful content.
– A safety evaluation system facilitates application testing pre-deployment, enhancing compliance and security readiness.

– **Microsoft Defender for Cloud**:
– Introduces AI security posture management, helping organizations to find and manage their AI inventories.
– Offers recommendations against cyber threats through continuous monitoring.
– Monitors for signs of cyberattacks, such as prompt injections and credential theft.

– **Integration of Security Alerts**:
– Alerts generated from security incidents are enriched with actionable evidence, aiding SOC analysts in comprehending user behaviors and attack vectors.
– Security activities related to generative AI applications are centralized in Microsoft Defender XDR.

– **Governance of Third-Party AI Apps**:
– Microsoft Defender for Cloud Apps allows organizations to assess risks associated with over 850 generative AI applications, providing insights into their security compliance and potential legal issues.
– It enables organizations to tag and block access to high-risk applications.

– **Data Security Measures**:
– Microsoft Purview Data Security Posture Management provides insights into compliance risks, particularly focusing on sensitive data in user prompts.
– A Data Loss Prevention policy can restrict elevated-risk users from sharing sensitive information with third-party AI applications, ensuring robust data governance.

– **Impact on Security Strategy**: The integration of these solutions significantly aids organizations in transforming their AI programs while maintaining a proactive stance against the evolving threat landscape.

The information provided emphasizes the necessity of continuous monitoring, risk assessment, and the integration of security considerations throughout AI application lifecycle management, ensuring compliance and governance in an increasingly AI-driven environment. These insights are invaluable for professionals working in security, compliance, and AI development, who must navigate the complexities of safeguarding AI workloads.

1 2 3 5 800 a access Act AI AI applications AI development ai model AI models AI safety AI security AI systems AI workloads alerts and Application applications ARM art as assessment attack attack vectors attacks Azure Behavior bert by C capabilities CIA Cloud compliance compliance and governance compliance risks content content safety continuous monitoring credential credential theft critical cyber cyber threat cyber threats cyberattack Cyberattacks D data data governance data loss data loss prevention Data Protection data security data security measures data security posture data security posture management de DeepSeek DeepSeek app DeepSeek R1 deployment design development driven e end Entra environment evaluation event exp filtering first for Foundry g Gen generated generative Generative AI git GitHub Go governance growth gs harmful content high Highlight HR http HTTPS in incident information injection injections insights integration IRS J k l land led Legal Legal Issues life lifecycle management low man management Micro Microsoft Microsoft Defender Microsoft Defender for Cloud Microsoft Purview Microsoft Security model models Monitor monitoring o of off on one OPM organization organizations out over party policy post posture management potential pre prevention proactive proactive stance professionals prompt prompt injections prompts R R1 rate RCE readiness recommendations red Risk Risk Assessment risks RMF Ro robust security RoT s safe safety Safety Evaluation sec secure secure environment security security compliance security considerations security incident security incidents security measure security measures security posture security risk security risks security strategy sensitive data sensitive information SHA sharing side Sig SoC source specific SSE Strategy system systems T test Testing text the theft third third-party threat threat landscape Threat Protection threats to tool tools Tor TP transformation UI US usage use user user behavior Users V val Valuation vectors visibility Wi workload workloads x