Inferensys

Glossary

Obligation Delta

The net change in a regulated entity's mandatory duties, prohibitions, or permissions resulting from an update to the governing legal text.
Stylish WeWork-like workspace with hot desks and document wall, professional searching through enterprise knowledge base on a mounted ultrawide display, warm industrial pendants overhead.
REGULATORY CHANGE DETECTION

What is Obligation Delta?

A precise measure of how a legal update alters an entity's mandatory duties, prohibitions, or permissions.

Obligation Delta is the net change in a regulated entity's mandatory duties, prohibitions, or permissions resulting from an update to the governing legal text. It quantifies the shift in the deontic landscape—what an organization must do, must not do, or may do—caused by a specific regulatory delta.

Computing an obligation delta requires mapping parsed textual amendments to a formal deontic logic model of the entity's compliance posture. A new statutory clause may create a positive obligation delta by adding a reporting duty, or a negative one by removing a prohibition. This metric is the core output of a compliance gap analysis, directly informing change impact scoring and triggering automated regulatory change workflows.

DEFINITIONAL COMPONENTS

Core Characteristics of an Obligation Delta

An Obligation Delta is not merely a textual diff; it is a structured, semantic analysis of how a legal update modifies the deontic landscape. The following cards dissect the core characteristics that define this net change in duties, prohibitions, and permissions.

01

Deontic Modal Transformation

The core of an Obligation Delta is the transformation of a deontic operator. It identifies when a permission becomes a prohibition, a duty is narrowed, or a new obligation is created.

  • SHALL → SHALL NOT: A mandatory duty is converted into a prohibition.
  • MAY → SHALL: An optional permission is elevated to a mandatory obligation.
  • Null → SHALL: A completely new affirmative duty is introduced.
  • SHALL → MAY: A mandatory duty is relaxed to a discretionary permission.
02

Actor-Scoped Applicability

The delta is always scoped to a specific regulated actor or class of entity. A single regulatory amendment can generate multiple, distinct Obligation Deltas for different stakeholders.

  • Regulated Entity: The party directly bound by the new duty (e.g., 'financial institution').
  • Supervisory Body: A new reporting duty imposed on a regulator.
  • Third Party: A new right or protection granted to a consumer or counterparty.
  • Mixed Impact: A single change that imposes a duty on one actor while creating a corresponding right for another.
03

Conditional Trigger Logic

An Obligation Delta captures the precise preconditions that activate the new duty. It models the logical 'if-then' structure of the rule, moving beyond simple text changes to semantic triggers.

  • Threshold Triggers: A duty activates only if a quantitative metric exceeds a limit (e.g., 'if assets > $50B').
  • Event Triggers: A duty is contingent on a specific event (e.g., 'upon a data breach').
  • Temporal Triggers: A duty becomes effective on a specific date or after a defined period.
  • Compound Triggers: Complex logic combining multiple conditions with AND/OR operators.
04

Operationalized Actionable Object

The delta defines the specific, actionable object of the obligation—what exactly must be done, prohibited, or permitted. This moves from legal text to a machine-readable compliance task.

  • Reporting Action: 'File Form X-17A-5 within 60 days.'
  • Prohibited Conduct: 'Shall not engage in proprietary trading.'
  • Policy Mandate: 'Must establish a written cybersecurity policy.'
  • Capital Requirement: 'Shall maintain a minimum Tier 1 capital ratio of 8%.'
05

Semantic Severity Weighting

Not all deltas are equal. A core characteristic is a severity score that quantifies the operational impact of the change, enabling triage and prioritization.

  • Critical (New Duty): A brand-new, affirmative obligation requiring immediate process creation.
  • High (Threshold Change): A modification to a numeric limit that alters compliance status.
  • Medium (Procedural Shift): A change to a filing method or timeline.
  • Low (Definitional Clarification): A non-substantive change that refines terminology without altering core duties.
06

Provenance and Grounding Link

Every Obligation Delta maintains a cryptographically verifiable provenance chain back to the specific source text. This ensures auditability and allows for precise legal citation.

  • Source Document: The specific amending bill or final rule.
  • Target Provision: The exact statutory section being modified (e.g., 15 U.S.C. § 78j).
  • Effective Date: The extracted date when the delta becomes operative.
  • Diff Evidence: The precise text span that constitutes the change, serving as the grounding for the semantic analysis.
OBLIGATION DELTA

Frequently Asked Questions

Explore the core concepts behind obligation delta analysis, the computational method for quantifying how regulatory amendments alter an entity's mandatory duties, prohibitions, and permissions.

An obligation delta is the net change in a regulated entity's mandatory duties, prohibitions, or permissions resulting from an update to the governing legal text. It is calculated by performing a deontic logic comparison between two versions of a regulatory instrument. The process involves: extracting deontic modalities (obligations, prohibitions, permissions) from both the source and target texts using legal NLP models; mapping each modality to a specific regulated actor and action; and then computing the set difference. The resulting delta categorizes each change as a new obligation, removed prohibition, modified permission, or similar transformation, providing a structured, machine-readable ledger of how the compliance landscape has shifted for a specific entity.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.