Microsoft says AI agents are like insider threats as ‘frontier models can’t be trusted’

Microsoft warns enterprises that AI agents could function as insider threats because frontier models cannot be fully trusted. The company advises organizations to closely monitor agents used by employees, especially those operating on systems not controlled by AI labs such as OpenAI and Anthropic. This emergence of AI agents as potential security risks necessitates the implementation of robust governance and oversight, which is currently reshaping enterprise security strategies to mitigate these specific internal vulnerabilities.