Defining a new protocol category for runtime governance of autonomous systems.

— 01

Execution Authority.

Most security architectures stop at who and whether. Identity is verified, access is granted or denied, and the system moves on. The harder question — and the one that defines this research — is what happens next.

Once an autonomous agent or service has been authorized to act, every operation it performs raises a separate set of questions. Was the action within the bounds intended at authorization time? Were the constraints evaluated and enforced in real time? Can an independent party reproduce the evidence that what happened was, in fact, what was permitted?

This research defines a protocol category — Execution Authority — that addresses runtime governance of autonomous machines. The work is focused on the post-quantum era, where assumptions about cryptographic durability and verifiability are being rewritten, and where the systems being governed are operating at a speed and scale that human oversight alone cannot keep pace with.

— 02

Adversarial governance integrity verification.

Governance mechanisms themselves can be attacked. An adversary that compromises the layer responsible for evaluating and recording authorization decisions can suppress, alter, or fabricate evidence — producing a record that looks correct but does not reflect what happened.

Adjacent research explores how the integrity of governance evidence can be verified under adversarial conditions, with constructions that hold even when the surrounding infrastructure is partially compromised. The intent is not to assume an honest substrate; it is to produce evidence that survives a hostile one.

— 03

The capability-governance gap.

Highly sophisticated autonomous AI systems are beginning to operate above the detection ceilings that conventional monitoring assumes. Logs, telemetry, and post-hoc review depend on the assumption that the monitored system is observable to the monitor. As capability advances, that assumption stops holding cleanly.

Future-oriented research at Yandeh examines this gap directly: how to design governance mechanisms that remain meaningful when the governed system is more capable than the governance layer itself, and what new constructions are required when traditional monitoring no longer suffices.