Governance Irony: Why Information-Heavy Compliance Directives Mechanically Fail in Large Language Models

Pour the complete rule-set into a system “to be covered” and adherence can collapse: with the relevant rule buried among competing directives, one language model’s compliance fell from 100% to 0% (another from 100% to 60–88%), while a short positive recipe held near 100% — length-, structure- and header-matched controls isolating the competing rules themselves as the cause. The paper argues that compliance is behaviour, not information, and that the dominant method — the monolithic rule-dump — is self-defeating. It uses a large language model as a measurable model organism: on the demands a rule-dump makes (attend to every clause, never tire, hold the whole manual at once) the machine is at least as favourably placed as a human reader, yet it still loses the buried rule, with human confounds (motivation, fatigue) absent — so the failure is structural, and capacity-graded (every model breaks once competing directives exceed its capacity). The fix is not less completeness but a different method: keep the complete record for audit, but install behaviour with a separate, short, retrieved, trained artefact (retrieval-augmented generation restores adherence to ~100% on a checkable task). The governance principle that follows: the completeness of a rule-record and the efficacy of the behaviour it installs are separable and routinely in tension; the target is not 100% adherence but optimal compliance, a calibrated point. Because the substrate makes a rule-set’s cognitive burden measurable, a language model can also serve as a regulatory-complexity test bed, pre-flighting a policy before it reaches people. The human-organisational extension is offered as a falsifiable hypothesis anchored in established human-factors literature (work-as-imagined / work-as-done; alert fatigue; the procedure paradox; information overload), not as a proof. Grounded in Behavioural Friction Theory (Paper 0) and its learning application, the Physics of Learning (Paper 16). v4 (August 2026) — attribution and scope revision. Retrieval-augmented generation is credited to Lewis et al. (2020) where the paper recommends it, and the abstract's claim for it is made conditional in the place where a reader would otherwise take it as unconditional. Restoring adherence to near 100 % rests on the retrieval finding the right clause: the method moves the burden from holding the whole rule-set to selecting from it, the selection was reliable on the tasks tested, and where triage fails the method inherits the failure it was meant to remove. Nothing in the experiments changes and no number moves. Earlier versions remain in the version history.