The most fascinating aspect about this sort of thing in my opinion is the realistic size limit of a "circle of trust". Considering the frequency of relationshi

The most fascinating aspect about this sort of thing in my opinion is the realistic size limit of a "circle of trust". Considering the frequency of relationship infidelity, I'd estimate the maximum reliable size of a circle of trust at roughly 1.5 people. Chances are intelligence agencies can use vetting procedures and internal processes to increase that number, but it's still fairly small. Many major scandals get discovered due to a broken circle of trust... via both professional and interpersonal infidelities. Human trust and trustworthiness is an act of loyalty and intimacy. Humans are notoriously easy to offend, challenging loyalties and disrupting intimacy. In the private sector, employees frequently jump ship to leverage trade secrets for their own benefit... not necessarily as blatant corporate espionage, but by having "several years of highly relevant industry experience" and applying for a job with a competitor. Something like Bitcoin requires us to trust any one party very little, but to trust the system's behavior in aggregate. The safeguards put in place after Manning and Snowden's breaches seem to be intended to mitigate harm, as does "need to know" security clearance, but I wonder if there exist (or will exist) more elaborate distributed trust systems: Suppose that instead of "need to know", everyone else in the intelligence agency department knew some information about what secret information you had viewed or saved that day, week, month... both magnitude and frequency, and frequency of repeated visits to document areas... some sort of footprint that offered an easy heuristic for social detection of deceitful behavior by fellow humans, rather than relying simply on harm-mitigation measures. This culprit would have had a very distinct footprint even if none of his coworkers could themselves access the data or see exactly what it contained. Disclaimer: I believe that government classification of information ought to be avoided in 95% of the circumstances it is used today. There are reasons to do it, but all information should be declassified within 50 years of its initial classification no matter what, and classification of information for political or selfish reasons (not directly in the national interest) should be a serious crime akin to treason or espionage.