Large language models often ignore regulatory rules in compliance decisions
Verdict Without the Rule: Diagnosing and Auditing Regulatory Rule Sensitivity in LLM Compliance Systems
Artificial IntelligenceComputation and LanguageComputers and Society
Summary
This paper finds that when AI models are asked to decide if something follows rules, their answers often don't change even if the rule they’re given changes a lot. The authors tested five popular models and found that these models sometimes ignore the specific rules and still give the same judgment. One specialized model meant to check rules was even less sensitive to rule changes and less accurate. This shows that just getting the right answer doesn’t mean the model really understands or uses the rule it’s supposed to follow.
What this means in practice
- •For compliance software engineers: Identify when compliance AI models ignore the provided rules to improve trustworthiness and debugging of regulatory decision systems.
- •For platform policy teams: Assess how policy enforcement tools rely on given rules versus default patterns to better design and audit content moderation systems.
Authors
Saisab Sadhu, Aadit Sengupta, Vinay kumar Sankarapu, Pratinav Seth
Abstract
Large language model compliance systems are deployed on the assumption that a verdict depends on the regulatory rule it is given. We test this directly across five models and 20 regulatory and platform-policy domains: delete, swap, or negate the governing rule while holding the case fixed, and check whether the verdict changes (OCS) or the model's internal representation of compliance shifts at all (ICS-delta). Neither moves much: models' verdicts are often invariant to substantial perturbations of the supplied rule, and the guard model, evaluated here under a custom-rule adaptation of its native taxonomy, is the least rule-sensitive and least accurate of the five, barely above chance (51%, versus 90-92% for general-purpose models). This reflects easy cases more than blanket neglect: on cases where deleting the rule changes a previously correct model prediction, models do track it closely. Neither better prompting nor direct intervention on the model's internal representations closes this gap. Accuracy alone does not establish that a compliance verdict is grounded in the supplied rule.