AI agents are beginning to behave less like tools and more like people. And like people, they can be misled. The difference is that the consequences can spread faster and further. This is not just a technical concern. It raises broader questions about trust and accountability in the systems shaping our daily lives. In this article, I share a personal perspective on why our security thinking may need to adapt.

L8insights is a personal space for reflection, where I write slowly to think things through and share my own perspective on systems, society and the quiet tension between them.