AI Agents Are Not the Problem. Our Security Posture Is.
Documented agentic AI failure modes, approval-gate weaknesses, supply chain compromise, and the forensic readiness needed to investigate them.
Research, analysis, and public writing
Analysis for subjects where certainty should be earned.
Research and analysis on digital forensics, cybersecurity, AI assurance, and the public interest.
Featured analysis
Read the analysis
Research and publications
Published method · Version 1.0
The Zemi Method
A published doctrine on investigative reasoning, evidentiary boundaries, and the disciplined separation of what is known, inferred, and unresolved.
Suggested citation: Watson, K. V. The Zemi Method, Version 1.0.
Visit The Zemi MethodWorking paper · July 2026
Artifact-to-Finding Promotion: Evidentiary Requirements for Harm from Correctly Functioning AI Systems
A proposed incident category and reconstruction test for cases in which a correct AI output is promoted beyond the warrant it establishes.
Suggested citation: Watson, K. V. (2026). Artifact-to-finding promotion: Evidentiary requirements for harm from correctly functioning AI systems (Working Paper v1.0). Zenodo. https://doi.org/10.5281/zenodo.21543521
Read the working paperAI assurance
Governance, testing, accountability, and the limits of AI-assisted decisions.
Digital evidence
Deepfakes, interpretation, forensic method, and what technical artifacts can reliably establish.
AI agent security
The security implications of increasingly autonomous systems and tool-enabled models.
Watson’s World
Jamaican and Caribbean history, crime, justice, politics, identity, and public memory.
Editorial method
What is known? What is inferred? What remains unresolved?
Those distinctions are not caveats added at the end. They are the structure of the work.