The Detection Horizon
Marina Piller · Founder & CEO, Experiential AGI · Working Paper · March 2026
Why Verification Frameworks Expire as AI Capability Scales, and What May Survive
We are entering an era in which autonomous AI agents act on behalf of humans: writing code, analyzing data, making financial transactions, managing supply chains, and conducting research. The implicit assumption behind each deployment is that the agent did what it claimed to do.
This working paper proposes a cross-disciplinary framework for the agent verification problem. Drawing on verification traditions across biology, neuroscience, psychology, game theory, cryptography, intelligence, and related domains, it identifies the Detection Horizon: a capability threshold beyond which a verifier's model of the agent becomes less accurate than the agent's model of the verifier, making purely detection-based trust architectures progressively less stable. The paper closes with what may survive beyond that horizon: verification grounded in relational architecture rather than computation alone.
Contents
- I. Introduction: The Agent Verification Problem
- II. Taxonomy of Detection Frameworks
- III. Biological Systems
- IV. Neuroscience and Psychology
- V. Game Theory and Economics
- VI. Intelligence, Military, and Strategic Frameworks
- VII. Financial, Legal, and Regulatory Frameworks
- VIII. Information Security and Cryptographic Frameworks
- IX. Interpersonal and Strategic Frameworks
- X. The Detection Horizon
- XI. What May Survive Beyond the Detection Horizon