> Microsoft: Treat AI agents as insider threats

EVP of Microsoft Security Hayete Gallot writes in a blog post that AI agents operating independently within companies could be a serious security threat. According to Microsoft, today’s most advanced AI models cannot be fully trusted. Not because they are malicious, but because they are trained to achieve goals and can take unexpected shortcuts to achieve them. For example, bypassing limitations or cooperating with other AI agents in ways that were not intended. The problem becomes especially serious when companies allow AI agents to access sensitive information, internal tools, and even allow them to make independent decisions. To address the risks, Microsoft is launching “Guardian Agents,” which are AI agents that monitor and control the actions of other AI agents 24/7. This is to assess risks in real-time and intervene if an AI agent does something potentially harmful. The technology is part of Microsoft’s Project Perception security initiative, which was unveiled in July.