Trending Topics
Trending
Microsoft releases security advice to keep Deepseek and AI tools safe on Azure AI Foundry
AI-driven software development is on the rise, as artificial intelligence and machine learning tools such as ChatGPT and DeepSeek continue to improve. As more companies adopt this technology, it will become increasingly important to take a security-focused approach during implementation. That’s why Microsoft recently announced the availability of DeepSeek R1 on Azure AI Foundry and GitHub.
And that’s what this article is all about: how to safeguard AI systems with Microsoft Security.
Azure AI Foundry security best practices
As a new addition to Azure AI Foundry, DeepSeek R1 has undergone a strenuous red teaming and safety evaluation. This includes model behaviour assessments and a slew of security reviews in an effort to reduce (or mitigate) risks and security threats. The end goal is for any AI models hosted within Microsoft’s Cloud service to be protected from data loss or theft while safely nestled in Azure’s secure walls.
For example, the Azure AI Content Safety tool provides automatic content filtering to detect and block content it deems malicious. For admins seeking more flexibility, an opt-out option allows administrators to set their own security level.
Another helpful add-on is the safety evaluation system, which lets developers test applications for security flaws prior to deployment. This helps ensure applications are safe, secure and compliant prior to launch.
Microsoft Defender for Azure AI Foundry
Microsoft Defender for Cloud’s AI security posture management tools are a valuable asset to security teams seeking to secure data – particularly those working with open source technologies.
The toolset lets admins discover and inventory AI models, orchestrators and data sources. They can also gain insight into associated workloads and detect AI cyberattack surfaces, pathways and other vulnerabilities.
The AI security tool then provides recommendations on how to strengthen security protocols to prevent such threats.
Application monitoring with Azure AI Foundry
While safeguarding applications and AI workloads, such as those built into DeepSeek R1, is a crucial first step, no amount of effort will stop all malicious actors.
So it’s important to use an application monitoring tool, such as Microsoft Defender for Cloud. Integrated with Azure AI Foundry, Defender continuously monitors DeepSeek AI apps and alerts security admins to threatening activity.
Further, the security feature provides invaluable insight into possible attacks by adding supporting evidence, helping security teams respond to jailbreaks, credential theft, data leaks and other acts.
Other data security concerns for AI within Azure
In addition to the above best practices and security features, the following services help secure AI apps created with DeepSeek and other AI app platforms hosted within Azure cloud services:
- Microsoft Purview Data Loss Prevention: Used to prevent and discourage users from copying and pasting sensitive information and uploading files that may contain sensitive data into Gen-AI apps. Customizable to target users who may pose an elevated risk, including external third-party applications.
- Third-party AI Governance: Microsoft Security is able to discover third-party AI apps used within your company and help admins govern their use, helping to bolster security and adhere to compliance standards.
- Risk Assessment: Microsoft Defender for Cloud Apps has built in risk assessment tools covering over 850 Generative AI apps. The tool is able to detect the use of this Gen AI software and assess them for their potential risk, security threat, compliance, and legal risk. Admins can then either tag them as “unsanctioned” or block user access, as appropriate.
In other news
About The Author
