Regulon.AI is an AI compliance intelligence platform built for product counsel and compliance professionals at companies that develop or deploy AI systems. It tracks state and federal legislation that imposes obligations on those companies — and maps each provision to a structured taxonomy of compliance requirements.
Most legislative trackers organize by bill. Regulon.AI's primary organizing unit is the compliance requirement — a discrete, actionable obligation type that recurs across multiple bills and jurisdictions. Each bill is broken down into one or more mappings: individual provisions extracted from the statutory text and assigned to a requirement in the taxonomy. Every mapping includes the verbatim statutory text, a plain-language explanation written for a legal professional audience, and the taxonomy ID linking it to every other bill that imposes the same type of obligation.
The result: a product counsel working on a disclosure audit can go to a single requirement page and see every bill in the corpus that imposes disclosure obligations — with verbatim statutory text from each jurisdiction, side by side. Regulon.AI is not a news digest, a policy scorecard, or a legal opinion service. It surfaces what the law actually requires, organized by obligation type.
Regulon.AI's primary legislative data source is LegiScan, a legislative tracking service that aggregates bill text, status, sponsor, and amendment data for all 50 U.S. states and Congress. LegiScan data is retrieved via API and refreshed continuously as bills are introduced, amended, or change status.
The following data points are pulled from LegiScan for each bill:
Candidate bills are identified by running keyword searches against the LegiScan monitoring API. Bills matching one or more keywords below are retrieved and evaluated for inclusion. A keyword match is necessary but not sufficient for a bill to be included — the full bill text is reviewed before any bill is added to the corpus.
| Group | Search Terms |
|---|---|
| Core AI | artificial intelligencemachine learningautomated decisionalgorithmicgenerative AIlarge language modelfoundation model |
| Output Types | synthetic mediadeepfakeAI-generated contentdigital replicavoice cloning |
| Application Contexts | automated employment decisionAI hiringbiometric surveillancefacial recognitionemotion detection |
| Governance | AI risk managementalgorithmic impact assessmentAI transparencyAI accountabilityAI safety evaluation |
| Model-Specific | training datacompute thresholdred-teamingmodel cardAI watermarkAI detection tool |
A bill is included in Regulon.AI if it satisfies both of the following criteria:
The bill must impose direct obligations on entities that develop or deploy AI systems. Bills addressing only consumer protection against downstream pricing practices, general civil rights with incidental AI language, or study commissions without operative obligations are excluded.
The bill must contain at least one provision that maps to a concrete, actionable compliance obligation in the requirement taxonomy. General legislative declarations of AI policy intent do not qualify on their own.
| Bill Type | Status | Reason |
|---|---|---|
| Algorithmic pricing prohibitions targeting downstream commerce | Excluded | Consumer-protection law targeting commercial conduct, not AI developer/deployer obligations. |
| AI study, task force, or advisory committee bills only | Excluded | No operative compliance obligations on regulated entities. |
| General civil rights bills with incidental AI language | Excluded | Obligation arises from existing civil rights framework, not an AI-specific compliance requirement. |
| Bills regulating only individual end users | Excluded | End users are the protected party in AI legislation. Regulon.AI tracks obligations on regulated parties. |
| Failed bills prior to 2022 | Excluded (currently) | Failed bill corpus from 2022 onward is planned as a future addition. |
Every obligation in Regulon.AI is assigned to a requirement — a discrete compliance obligation type with a stable identifier. Requirements are organized into 11 categories covering the full range of obligations AI legislation currently imposes on developers and deployers.
The taxonomy is the backbone of the cross-bill structure. A requirement page for, say, AI Identity Disclosure links to every bill in the corpus that imposes that obligation — letting teams understand how the requirement varies in scope, threshold, and applicability across jurisdictions without reading each bill individually.
The Passage Likelihood indicator is a rule-based High / Medium / Low signal displayed on bill detail pages for active pending U.S. bills. It is intended to help compliance teams prioritize which bills warrant close monitoring and early planning.
The indicator is derived entirely from structured data returned by the LegiScan API. No editorial judgment is applied. It is not shown on enacted laws or failed bills.
Scoring is evaluated top-down: each bill is assessed for High first, then Medium, then Low. A bill that qualifies as High is not re-evaluated against Medium criteria.
Rule-based, not predictive. The indicator applies deterministic rules to structured data. It does not use statistical modeling, historical passage rates, or machine learning. Political dynamics, leadership opposition, and budget constraints are not captured.
Committee passage detection is imperfect. LegiScan history action strings are free-text and vary across states. Keyword matching may miss atypical phrasings used in some legislatures. Missed detections are treated conservatively — a bill with an undetected committee passage may show a lower signal than warranted.
Federal bills. The indicator applies to U.S. state and federal bills where majority party data is available. Bills from jurisdictions absent from the majority party lookup will show an indeterminate sponsor-party signal.