I’m advising with Lasting Asset at the current time on bringing digital trust to agentic AI, and building new methods of trust that would allow our agents to have a proper digital ID and that can be linked to a person and/or role. This will better support the tracing of fraudulent activities. And, over the past few weeks, I’ve been presenting my doom-laden view of the changing face of cybersecurity at a range of conferences.
For this, I personally believe that the move to agentic AI will completely change the world of cyberecurity and we need to make a move now to avoid worst-case scenarios. Concepts such as , “humans-in-the-loops”, “kill-switch architectures” and “zero trust for agents” have hardly been talked about in the conference events I’ve been to, and where the industry just keeps focusing on the same old issues — could we be sleepwalking into an agentic AI nightmare?
Imagine you leave a whole lot of young children on their own in a kitchen and ask them to bake an apple pie. Can you imagine the mess you would come home to? Well, welcome to the world of agentic AI.
And, so, a new paper from Stanford and Harvard blows a hole in the cybersecurity [here]:
