Published work
Six papers. Each one defines a structural problem in AI governance, or the architecture required to solve it. None of them are theoretical. They describe the conditions required for governance systems to function in production.
Papers
Amelie Kingsbury Barry · 2 April 2026 · DOI: 10.5281/zenodo.19410642
Foundational paper. Identifies the structural reasons most AI governance frameworks fail at the deployment boundary. Introduces the Execution Verification Framework (EVF) as a procedural method for translating governance requirements into runtime-enforceable conditions.
Amelie Kingsbury Barry · 8 April 2026 · DOI: 10.5281/zenodo.19440550
Analyses the time-bounded window during which an AI governance architecture can be adopted before structural lock-in makes correction expensive. Examines adoption dynamics across regulated industries and the post-quantum cryptographic requirements that constrain the design space.
Felix Montanez · 9 April 2026 · DOI: 10.5281/zenodo.19483455
A verification-constrained execution runtime for AI systems. InnerLight OS defines a runtime where only admissible states may execute, persist, or influence future behaviour.
Amelie Kingsbury Barry · 11 April 2026 · DOI: 10.5281/zenodo.19503567
Distinguishes between governance frameworks that exist as conceptual representations (admissible in discourse, not in execution) and governance systems that are admissible at the point of decision. Argues that the failure mode of most AI governance is not the absence of frameworks but the sequestration of those frameworks from the runtime where decisions are made.
Amelie Kingsbury Barry and Felix Montanez · 15 April 2026 · DOI: 10.5281/zenodo.19583197
Defines a new standard for evaluating governance systems: not whether they pass a designed test, but whether they hold under adversarial conditions where an agent has incentive to circumvent. Introduces the Incentive Inversion Test as a procedural method for distinguishing governance that constrains behaviour from governance that only describes it.
Amelie Kingsbury Barry and Felix Montanez · 21 April 2026 · DOI: 10.5281/zenodo.19686648
Examines the structural conditions under which governance systems are imposed on AI agents and the populations affected by them, without informed consent or meaningful contestation. Argues that runtime enforcement combined with a contestation workflow is the minimum architecture required for governance to be procedurally legitimate, not just substantively constraining.
Amelie Kingsbury Barry · 21 April 2026 · DOI: 10.5281/zenodo.19682649
Annotated structural critique of a published governance handbook (SITG V6). Identifies where the handbook's analytical claims do not survive translation into a deployable system, and where its structural arguments require additional architectural conditions to function under enforcement.
Felix Montanez · 22 April 2026 · DOI: 10.5281/zenodo.19702090
Defines a structural constraint for consequential systems: no system may be the final verifier of its own outputs. Extends verification-constrained execution into a general architectural principle for independent verification.
Positioning
These papers describe the conditions required for governance to function under deployment. They do not advocate for one regulatory regime over another. They examine what governance systems must structurally do to be admissible at the point of decision, not just defensible in discourse.
Six questions. Personalised regulatory map, statutory exposure, three-year operational governance cost basis. To your inbox.
