Yeah, I just

wanted to add.

So you make absolutely that that is one of the sort of key areas that we need to be aware of.

But the other thing is just general awareness for everybody.

There's this concept called the lethal trifecta, which was coined in June of 2025, so it's only a year old, by Simon Willison.

And it's essentially the idea that if you give an agentic AI, so this is an agentic AI, the ability to leave the organization externally, to have access to sensitive data, and permissions to action itself, or to do if you're giving it a particular thing, you can inadvertently create a hole that your security perimeter will not be able to close, right? Because the information that it's processing is untrusted.

And by untrusted, I mean from anybody.

So do we allow them to shut down a workstation but not the core? These are the men to ask,