I was thinking more about the soft targets and people based parts of the system, but you raise an interesting point. Zero trust systems might be solvable from a tech perspective, but zero trust organizations from a social perspective are a whole lot harder to design and enforce, maybe impossible.

In accounting systems, which are targets of fraud and malfeasance (e.g. Enron), there is a "4 eyes" principle. It takes at least 2 people to change something that impacts the financials. But that can't stop 2 people from colluding.

Back to the topic, if one agent or informant builds trust with one actor in the target group, and they collude or if one slip is made, it could be game over. Once the org is compromised, how do you know who to trust?