{"id":1597,"date":"2025-11-25T08:00:05","date_gmt":"2025-11-25T16:00:05","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/foundry\/?p=1597"},"modified":"2025-12-08T09:50:11","modified_gmt":"2025-12-08T17:50:11","slug":"assess-agentic-risks-with-the-ai-red-teaming-agent-in-microsoft-foundry","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/foundry\/assess-agentic-risks-with-the-ai-red-teaming-agent-in-microsoft-foundry\/","title":{"rendered":"Assess Agentic Risks with the AI Red Teaming Agent in Microsoft Foundry"},"content":{"rendered":"

We\u2019re thrilled to announce major enhancements to the AI Red Teaming Agent<\/a> in Microsoft Foundry for models and AI agentic pipelines, available now in public preview<\/strong>. These new capabilities enable organizations to proactively identify safety and security risks in both models and agentic systems, ensuring strong safeguards as agentic solutions move into production workflows.<\/p>\n

The AI Red Teaming Agent integrates Microsoft AI Red Teaming team\u2019s open-source framework, <\/strong>PyRIT (Python Risk Identification Tool)<\/strong><\/a> to deliver automated, scalable adversarial testing. With PyRIT\u2019s reusable attacker strategies and orchestration capabilities, teams can systematically probe for vulnerabilities and risks such as prompt injection attacks, harmful content generation, misuse enablement, privacy leaks, and robustness failures, utilizing more than 20 attack strategies. This integration makes red teaming more consistent, data driven, and reproducible, thereby helping organizations move from ad-hoc testing to structured AI safety and security assessments. In this post, we\u2019ll highlight the key innovations brought to you by Microsoft Foundry and provide a technical dive into how the AI Red Teaming Agent can help you tackle agentic risks that are top of mind for many organizations.<\/p>\n

Foundry as a unified platform for automated red teaming for models and agents<\/strong><\/h2>\n

Organizations and teams can now orchestrate automated red teaming runs across both model-level and agent-level surfaces through the unified Foundry SDK\/APIs and UI portal. This means you can test not just large language models or systems alone, but also in more complex, tool-using agents, enabling end-to-end risk coverage in both agentic and non-agentic scenarios.<\/p>\n

No-code UI wizard:<\/strong> Run automated red teaming runs without writing a single line of code. This is ideal for testing during rapid prototyping and enables non-technical stakeholders to also run quick red teaming runs for compliance and governance.<\/p>\n

\"1<\/a><\/p>\n

View and compare red teaming results:<\/strong> Foundry\u2019s new interface for evaluations lets you analyze risk profiles across generative AI models and systems (agentic and non-agentic), track safety vulnerabilities, and benchmark improvements over time, all in one place.<\/p>\n

\"2<\/a><\/p>\n

Foundry SDK & APIs for batch and continuous red teaming:<\/strong> Integrate red teaming into your CI\/CD pipelines, schedule and kick off red teaming runs remotely for continuous agent safety evaluation<\/strong>.<\/p>\n

With these new functionalities in Foundry, the AI Red Teaming Agent helps developers start their Trustworthy AI journey, providing:<\/p>\n

    \n
  • End-to-end risk coverage:<\/strong> Test your agents in simulated adversarial scenarios that reflect potential deployment risks.<\/li>\n
  • Dynamic and customizable:<\/strong> Dynamic seed dataset generation based on relevant risk definitions and policies, selected attack strategies, agent tool use and evaluation criteria.<\/li>\n
  • Continuous integration:<\/strong> Automate red teaming as part of your development lifecycle to catch regressions and emerging risks early.<\/li>\n<\/ul>\n

    The AI Red Teaming Agent is helping customers like Blue Yonder assess and catch generative AI risks before deploying to production. <\/strong>Watch the full video testimonial here<\/a>.<\/p>\n

    \u201cFoundry’s AI Red Teaming Agent helps us proactively identify security risks, enabling us to mitigate issues before they become costly safety incidents.”<\/span><\/p>\n

    Chris Burchett, Sr. Vice President Generative AI, Blue Yonder<\/strong><\/p><\/blockquote>\n