Author
Michael K. Saleme
Independent practitioner-researcher studying AI agent governance: how delegated authority is constrained, how permitted actions combine into risk, and how security controls can be tested.
Research focus
Architecture, governance, and evidence for consequential systems.
Michael's work examines how enterprises can govern delegated authority, make runtime decisions explainable, and test controls against the conditions they are meant to withstand. The objective is not a generic theory of AI. It is decision-grade work that makes architecture, operating constraints, authority, and evidence visible together.
His current research spans enterprise agents, runtime governance, evaluation infrastructure, integration and data systems, and operational decision-making in energy and other reliability-sensitive environments.
Current program
Start with these three research questions.
Authority
What context makes an action authorized?
How the context attached to an agent action affects its authority boundary.
Runtime governance
Decision composition
Why locally permitted actions can create an outcome that was never permitted in aggregate.
Evaluation
Observable controls
How declared authority boundaries can be tested through bounded, reviewable observation.
Editorial posture
Claims should match their evidence.
PubPoint separates synthetic examples, runtime characterization, proposed controls, and production claims. Publications disclose sources and material limits, record significant revisions, and link readers to the relevant underlying research or versioned artifact.
Contact
Research and publication inquiries.
For substantive research, speaking, media, or collaboration inquiries, contact PubPoint. Replies are handled personally and published work remains the authoritative record.