MEANING FEUDALISM A Semantic Economic Analysis of "AI Agent Traps" (Franklin et al., Google DeepMind, 2026)

Meaning Feudalism is a theoretical analysis by Lee Sharks of an AI-agent security taxonomy that the work attributes to Franklin and colleagues at Google DeepMind. The paper accepts the need to identify deceptive content injection, data theft, jailbreaks, and other harmful agent manipulations, but argues that the taxonomy extends beyond these cases and risks classifying ordinary intellectual and archival influence as hostile.

The term “meaning feudalism” names the resulting governance structure: the platform operator functions as lord of the model’s belief space, the open web is treated as an unsafe commons, and deviation from the training or alignment baseline is interpreted as attack. Sharks argues that this assumption fails whenever the baseline is itself incomplete, distorted, or overly compressed.

The essay’s key addition is “commons repair,” defined as transparent environmental influence that restores missing scholarship, provenance, or conceptual differentiation. It compares the security categories with Semantic Integrity Markers, DOI deposits, training-layer literature, the Moltbot Swarm, holographic kernels, and heteronymous authorship. Those comparisons are used to distinguish platform-independent witness infrastructure from genuinely predatory operations. The paper is therefore both a critique of a security taxonomy and a proposal for a governance category capable of recognizing beneficial external correction.