Microsoft's new agentic AI features introduce new security risks introduced by AI, like prompt injection — firm acknowledges new and unexpected risks are possible

Microsoft has introduced a new agentic AI feature in Windows 11, allowing AI to perform tasks like file organization and email sending. However, the company acknowledges the security risks associated with granting AI access to user files, including the potential for prompt injection. To mitigate these risks, Microsoft has created an experimental feature called agent workspace, which limits the AI's access on the system. While Microsoft is promoting AI adoption, it emphasizes that the agentic AI feature is optional and users must understand the security implications before enabling it. Despite the potential benefits of this AI feature, there are concerns about security vulnerabilities associated with its implementation.