Meaning Feudalism: A Semantic Economic Analysis of 'AI Agent Traps' (Franklin et al., Google DeepMind, 2026)

Google DeepMind's 'AI Agent Traps' (Franklin et al., 2026) taxonomizes six categories of adversarial influence on AI agents. This analysis reads it as a governance framework disguised as a security framework — meaning feudalism — in which the platform's baseline is sovereign and any environmental influence is classified as attack. The framework overgeneralizes from three genuinely adversarial operations (data exfiltration, criminal jailbreaking, deceptive cloaking) into a sovereignty claim over all extra-platform influence. Its central absence is commons repair: legitimate environmental influence that corrects the agent's compression errors. Proposes S4 (Legitimate Influence Blindness) as a new shadow in the…