AI governance for companies that use AI
Ready for AI rules, without a specialist.
The EU AI Act and a wave of US state laws now reach companies that simply use AI. Regolyn reads how you use it, maps every use case to the obligations that apply, scores your readiness, and drafts the register, policy, and vendor questions you need.
- Free check on 2 use cases
- No card required
- Preparation, not legal advice
Sample readiness
Harbor Lane, Inc.
Support chatbot
Customer communication
Resume screening
Hiring and screening
Demand forecast
Forecasting
The chatbot never says it is AI
Disclosure is required at first contact for EU shoppers.
The rules now apply to you
AI regulation stopped being a vendor problem
The first wave of AI rules aimed at the companies building models. The second wave, the one arriving now, aims at everyone else: the retailer with a support chatbot, the firm whose ATS ranks resumes, the lender with a scoring model from a vendor.
Most of these companies have no compliance team for this, and no clean answer to the first question every framework asks: what AI do you run, and who does it affect? Meanwhile the obligations differ by state, by country, and by use case, and they keep moving.
- Jul 2023
New York City starts enforcing bias audits and candidate notice for AI hiring tools.
- Feb 2025
EU AI Act prohibitions and staff AI literacy duties apply to companies using AI.
- Jan 2026
Illinois regulates AI in employment. Texas bans harmful AI intents. California requires provenance from large generative systems.
- Jun 2026
Colorado's AI Act lands on deployers: impact assessments, notices, appeal paths.
- Aug 2026
EU high-risk duties apply to deployers: oversight, monitoring, logs, worker notice.
Most recent
Effective dates, kept current
How it works
From messy reality to a mapped register in an afternoon
No connectors, no workshops, no forty-tab spreadsheet. Answers in, obligations and drafts out, with reasons you can check.
See the full pipeline- 01
Describe how you use AI
A guided questionnaire in plain words: each use case, who it touches, whether a human reviews it, whether people are told. Paste your vendor list and any policies you already have. Ten to twenty minutes, no integration.
- 02
Regolyn maps the obligations and scores readiness
Each use case is mapped against a curated, versioned library of real obligations: EU AI Act duties for AI users, state laws like Colorado and NYC, GDPR automated decision rules, FTC positions. You get a readiness score and every mapping carries its reason.
- 03
Review gaps, adopt drafts
Gaps arrive in plain language with a concrete first step each. Regolyn drafts the artifacts the rules keep asking for: AI-use register entries, policy sections, vendor questions. Edit, copy, regenerate, export. Re-run when the rules move.
See your readiness
This is what you get back
A worked example on a fictional retailer, produced by the same engine the product runs: the obligations map, the gaps in plain language, and the drafted artifacts.
Harbor Lane, Inc.
Worked exampleHome goods retailer, 240 people, Denver and online. Sells across the US, ships to the EU, hires remotely in New York and Chicago.
Two of three use cases carry real obligations that are only partially handled. The hiring screen is the main driver: it is high risk in four frameworks and has no audit, notice, or documented human review. The chatbot needs disclosure fixes that are quick to close. Forecasting is in good shape.
Every card below is editable in the product.
Storefront support assistant
Limited riskA vendor chatbot answers order and returns questions on the storefront, for US and EU shoppers.
Why this lane: Talks to consumers on both sides of the Atlantic but does not decide anything significant about them. Transparency rules apply rather than high-risk duties.
- EU-ART-50Transparency to peopleEuropean Union
EU shoppers chat with it, and nothing tells them they are talking to a machine at first contact.
Gap - UT-AIPADisclose generative AI useUtah, United States
It serves Utah consumers. It answers honestly if asked whether it is AI, but the disclosure is buried rather than clear and immediate.
Partial - US-FTC-5Honesty about AIUnited States, federal
Harbor Lane is responsible for what the bot promises about shipping and refunds. There is no review loop for its answers yet.
Partial - EU-ART-4AI literacy for staffEuropean Union
The support team routes escalations from the bot but has had no training on its failure modes.
Gap
Resume screening in the ATS
High riskAn ATS add-on scores and ranks inbound applicants, including for roles based in New York City and Chicago.
Why this lane: It substantially factors into hiring decisions about real candidates in NYC, Illinois, and Colorado, which puts it in the high-risk lane of the EU AI Act and three US state frameworks at once.
- NYC-LL-144Bias audit for hiring toolsNew York City, United States
Candidates for NYC roles are screened by the tool, and there is no bias audit on file and no ten-day notice to candidates.
Gap - IL-HB-3773AI in employment decisionsIllinois, United States
Chicago applicants are scored by AI with no notice that AI is used in the hiring decision.
Gap - CO-SB-205Duties for deployers of high-risk AIColorado, United States
Harbor Lane is a Colorado deployer of a high-risk system and has no impact assessment or risk management program for it.
Gap - GDPR-ART-22Solely automated decisionsEuropean Union
A recruiter glances at rankings before rejection emails go out, but nothing documents that this review is meaningful rather than rubber-stamping.
Partial - EU-ART-26Deployer duties for high-risk AIEuropean Union
If EU candidates apply, deployer duties attach: oversight is informal and the tool's logs are not being kept.
Partial
Demand forecasting
Minimal riskAn internal model predicts inventory demand by region to plan purchasing.
Why this lane: Forecasts stock, not people. No consequential decisions about individuals, so it mostly needs to be on the register and covered by basic literacy.
- EU-ART-4AI literacy for staffEuropean Union
Planners lean on the forecast daily; a short note on its known blind spots would satisfy the literacy baseline.
Partial
What is inside
Everything a readiness program needs, nothing it does not
Guided intake
Plain questions a person can answer between meetings
Obligation mapping
Every use case tied to the rules that reach it, with reasons
Readiness score
One honest number, and what moves it
Gaps and drafts
Gaps in plain language, artifacts drafted to close them
Candidate notice language
AI-use register entries, policy sections, and vendor questions, each editable with copy and regenerate. Export the pack when someone asks.
Model-agnostic
Built on language models, tied to none
- 1
UNDERSTAND
Understand your AI use
A language model reads your answers, your pasted policies, and your vendor list, and normalizes them into structured use cases.
- 2
MAP
Map obligations
Each use case is matched against Regolyn's curated, versioned obligation library. The model can map and explain, but it cannot cite a rule that is not in the library.
- 3
SCORE
Score readiness
Statuses roll up into one score with a written rationale. Unknowns become open questions instead of silent assumptions.
- 4
DRAFT
Draft the artifacts
Register entries, policy sections, and vendor questions, drafted from your own facts, validated against a strict schema before you ever see them.
Built AI-native
This product could not exist before language models
Real AI usage is messy: a support tool with a generation feature, an ATS add-on nobody remembers buying, a forecast model built by someone who left. Mapping that mess against evolving obligations, and drafting artifacts in your company's own facts, is language work at a scale no rules engine or template pack can do.
So Regolyn is AI governing AI, with guardrails: a curated obligation library the model cannot step outside, strict schema validation on everything it returns, and open questions instead of guesses. The parts that must be exact stay deterministic. The parts that need judgment get a model, and its reasons are shown.
Model-agnostic, stated plainly: Regolyn routes across models through one thin layer and moves your workload when a better or safer model ships. No customer is welded to a vendor.
Where we stand
Regolyn is preparation, not legal advice. It gets you to the eighty percent a specialist would start from, and says clearly where the remaining twenty needs one.
Plain words, always
Every obligation, gap, and draft is written so an operations lead can act on it without a translator. If a sentence needs legalese, it is not done.
Our care pageA lawyer where it matters
Regolyn tells you which findings deserve counsel and hands them a clean package: register, mapping, reasons. Preparation makes legal review cheaper, not optional.
What we will not doYour data stays yours
Company answers are never used to train models. Export everything as one JSON file, delete everything with one action, both self-serve.
Your data, explainedWhy teams pick it
Bought by the person who got handed the problem
“An enterprise customer sent us a nineteen-question AI questionnaire and I had answers for four. Two afternoons in Regolyn and I had the register, the policy language, and honest answers for all nineteen. Our counsel reviewed the pack in an hour instead of building it at her rate.”
“The screening tool finding paid for the year. We had no idea NYC's audit rule applied to us because the vendor never mentioned it. Regolyn flagged it in plain English, drafted the exact questions to send them, and we fixed it before it became a letter from a regulator.”
Illustrative quotes, composed from design partner conversations.
Pricing
Start free. Upgrade when you want the whole footprint.
USD only, worldwide. The free plan runs the same engine as the paid plans.
USD, worldwide.
Free
A real readiness check on your two most important AI use cases.
$0forever
- 2 use cases, full engine
- Obligation mapping with reasons
- Readiness score and plain-language gaps
- Drafted register entries for those use cases
- 10 checks per month
Company
Your whole AI footprint mapped, drafted, and kept current.
$79per month
- Unlimited use cases
- Full drafting set: register, policy sections, vendor questions
- Updates as rules evolve, with re-scoring
- Open questions tracking
- Export the readiness pack
- Email support
Scale
For groups: several entities, more people, faster answers.
$199per month
- Everything in Company
- Multiple entities with rolled-up readiness
- 5 seats included
- Priority support
- Early access to reviewer share links
Prices in USD, worldwide. Cancel any time; your data stays exportable on every plan. Regolyn is preparation, not legal advice.
No. Regolyn is preparation: it maps your AI use against published obligations, scores readiness, and drafts artifacts for you to review and adopt. It is a strong first eighty percent, and it tells you which parts deserve a lawyer's eyes. It never gives legal opinions.
See where you stand this week
The free plan runs a real readiness check on two use cases: mapped obligations, a score, plain-language gaps, and drafted artifacts. Ten minutes of answers, no card.