Feature Overview
AI Guardrails for Microsoft Copilot is an expansion of our existing AI Guardrails—which already support various AI tools including GitHub Copilot—that delivers oversight for chat interactions. It focuses on safeguarding personally identifiable information (PII) and ensuring responsible, ethical use of generative AI tools in the workplace.
Key Benefits / Value Delivered
-
Safeguard Sensitive Information: Prevent the exposure of personally identifiable information (PII) in Microsoft Copilot chat by applying advanced privacy guardrail designed to protect confidential data.
-
Ensure Safe and Ethical Usage: Enforce comprehensive safety and security guardrails that prevent attempts to circumvent internal AI restrictions and actively block unethical behaviors, such as toxic speech or other inappropriate content, within Microsoft Copilot interactions.
Getting Started
To begin using AI Guardrails for Microsoft Copilot:
-
Go to the Data Loss Prevention Policy page under the Security menu in Cisco Secure Access.
-
Create a new AI Guardrail rule or edit an existing one.
-
In the Destinations section/criteria, select Microsoft Copilot and choose whether to apply the guardrail to prompts, responses, or both.
Documentation and Resources
Best Practices
-
Regularly review and update AI Guardrails policies as new use cases for AI chat emerge within your organization.
-
Monitor policy violation reports to identify patterns and refine enforcement rules.
-
Educate end users on responsible AI chat usage and the importance of data privacy to promote compliance.
-
Customize end user notifications to provide actionable guidance when a chat is blocked, supporting a positive and transparent user experience.
