Constitutional Epistemology Under Adversarial Semantic Pressure

This paper examines adversarial semantic pressure: behavior that exploits the representational and procedural forms of governance so a system appears governed while its claims, effects, authority, completeness, or semantic commitments remain false or ungrounded. Deliberate deception is one cause, but the same observable failure can arise from proxy optimization, institutional incentives, automation bias, distribution shift, or self-consistent error. The definition is therefore behavioral before it is psychological. The paper analyzes layer laundering, indefinite epistemic debt, typed-refusal manipulation, semantic drift under formal compliance, and the broader threats of selective omission, forked histories, compromised keys, evidence poisoning, observer capture, replay, and evaluation gaming. It distinguishes what receipt syntax, hash linkage, signatures, independent checkpoints, external effect observations, and completeness sampling can each establish. A receipt chain alone supports a bounded claim about the structure of a declared narrative; it does not prove wall-clock history, actor identity beyond key control, effect occurrence, semantic accuracy, or completeness. The proposed EI controls are recast as epistemic defense obligations rather than logically necessary or sufficient invariants. Different realizations may satisfy the same objectives through different mechanisms. Their function is to bind constitutions, semantic vocabularies, debt budgets, witnessed checkpoints, behavioral observations, key custody, and anti-equivocation evidence to independent trust domains. Within the current corpus architecture, the Governed Meaning Base (GMB) preserves governed semantic state; the Three Hash Rule separates state, admissibility, and meaning commitments; Cryptographic Phase binds those commitments to declared epochs; Semantic Custody preserves source and transformation history; and Reasoning by Way of Receipts exposes intermediate epistemic motion. These mechanisms increase the cost and detectability of deception under stated assumptions. They do not guarantee that an adversary's optimal strategy converges on honesty; that conclusion requires an explicit utility, detection, sanction, and capture model.