Guide · Published November 14, 2026 · Reviewed November 14, 2026 · Ingenious AI team · Not legal advice
The Free ADMT Duties Checker: How to Read Your Result Page
The short answer: The free duties checker evaluates three answers — jurisdiction (Colorado, California, Connecticut, European Union), decision type (employment, housing, credit/lending, education, other), and the AI's role (makes the decision alone, a substantial factor, or a minor input) — and returns a structured result: a verdict (duties apply / no duties found / check the list), a headline and summary, a date line, the duties with their effective dates and in-force status, notes carrying the enforcement and edge-case facts, the sources each jurisdiction's data was verified against, and the not-legal-advice disclaimer. This guide explains how to read each element — and what to do with the result.
Check your duties and dates — free Build your first notice — free
What is the checker? It is the public evaluation built on app/checker_data.py — duty data verified against primary or authoritative sources on October 28, 2026 (statute text, official bill pages, the CPPA regulations' text, EUR-Lex materials), with sources listed per jurisdiction on the result. It gives general information about published duties; it is not legal advice, and its disclaimer says so on every result.
The verdict line: three answers, three meanings
'Duties apply' means the combination you described is covered: the duties listed are yours, on the dates shown. 'No duties found' is always conditional on the answers — it means this role answer (a genuinely minor input, a genuinely human decision) takes the use outside the ADMT frame in that jurisdiction; if the role answer was optimistic, the verdict is worthless, which is why the guides insist on answering the role question from measured workflow facts (override rates, outage tests) rather than from the workflow diagram.
'Check the list' is the conditional verdict for the 'other' decision type: the checker cannot tell whether your decision is on the jurisdiction's statutory list, so it shows the list and the duties that would apply if it is. Treat a conditional verdict as a research task with the list in front of you — the comparison guide works through the three lists — not as a soft no.
The duties list: dates and status are the content
Each duty arrives with its effective date and a computed status — in force or upcoming — against the current date. Read the list as a calendar, not a catalogue: the in-force items (Connecticut's trafficking of dates is the extreme case, with duties live since 2023 alongside 2027 notices) are obligations you may already be behind on, while upcoming items carry build deadlines. The summary's date line compresses the same information for the jurisdiction as a whole.
Duty wording on the result is deliberately close to the sources — 'at least two easy ways to opt out', 'within 45 days (one 45-day extension is allowed)' — because operational teams build from these sentences. Where a duty's wording in the checker differs from a summary you have read elsewhere (including in the kit's own rules data, compiled from secondary summaries), the checker's verified data is the one to build from.
Worked example: Fairhaven Consumer Finance
Example — fictional business, real duty structure. Fairhaven, the fictional lender from the opt-out logging guide, used the checker as the first step of its California programme. Its session, combination by combination:
- Combination one: California + lending + substantial factor → 'Duties apply': pre-use notice, opt-out, access, risk assessment, all dated January 1, 2027 — the programme's California duty set, filed to the credit tool's register row.
- Combination two: California + lending + minor input (for its document-automation tool) → 'No duties found', with the human-decision reasoning stated — filed as the documented negative, with the override-rate evidence that supports the role answer.
- Combination three: Connecticut + housing + substantial factor (for a partner workflow) → duties already apply — the opt-out and outcome rights live since 2023/2026, which re-sequences that work ahead of the 2027 builds.
- Notes harvested: the California threshold statement prompts the coverage file (met via consumer volume); the Colorado AG-rules note is diarised for its lending partner's state.
- Sources followed: the CPPA regulation sections cited on the result become the programme's reading list; the whole session's outputs are filed, dated, in the register.
The notes: where the edge cases live
The notes section carries the facts that decide real programmes: threshold statements (California's business thresholds; Connecticut's 35,000), enforcement frames (Attorney-General enforcement, private-right-of-action positions, cure periods), scope exclusions (Connecticut's scheduling and health-and-safety carve-outs), and pending-items (Colorado's Attorney General rules on post-decision disclosures, due by January 1, 2027, which may change the explanation's exact format).
Read the notes before designing anything. Teams that read only the duties list build the centre of the programme and miss the threshold question that decides whether they are in it, or the exclusion that takes a system out. The notes are also where the checker is most honest about its own limits — the GDPR pointer sits there for the EU, labelled a different regime and outside the checker.
The sources: why they are printed on the result
Every jurisdiction's result lists its sources — SB 26-189 via the Colorado General Assembly, the CPPA regulations by section, Connecticut's Public Acts 26-15 and 25-113 with their General Statutes sections, the EU's Regulations 2024/1689 and 2026/1744 via EUR-Lex — with the verification date (October 28, 2026). That printing is a design decision: a duty claim that cannot name its source is an assertion, and the checker's authority comes from refusing to assert.
Use the sources as the reading list for anything the programme will spend real money on, and re-check them when the calendar says to: rules adoptions (Colorado's, due January 1, 2027) and amendments are exactly the events that stale a verified dataset. The kit's guides restate checker data; where a guide and a future source disagree, the source wins, and the guides in this series are written to be corrected by it.
After the result: from verdict to programme
File the result. The checker output for each combination — jurisdiction, decision, role — belongs in the register row for that tool, dated, as the duty-set evidence the rest of the programme builds on. Then follow the dependency order this series uses: assessments where owed, notices placed and generated, response workflows staffed, the record underneath. Each guide in the series starts where a checker result ends.
Re-run on change: a new tool, a new state, a changed role (the tool's weight in the workflow grew), or a source update. Verdicts are dated findings about described facts, not permanent classifications. And keep the disclaimer in view: the checker gives general information about published duties from verified sources — for a coverage decision with real money behind it, the result is the starting document for advice, not a substitute for it.
- Answer the role question from measured facts — override rates and the outage test — never from the workflow diagram
- Read the verdict precisely: 'duties apply' builds, 'no duties found' is conditional on the role answer, 'check the list' is a research task, not a no
- Read the notes before designing: thresholds, enforcement frames, exclusions, and pending items all live there
- File every result, dated, to the tool's register row — and re-run when the tool, the state set, the role, or the sources change
- Follow the sources for anything expensive, and treat the result as the starting document for advice, not a substitute for it
Sources, as carried in the kit's own data: the duty summaries in app/rules.py and the checker data in app/checker_data.py, verified against primary or authoritative sources on October 28, 2026. Run the free duties checker for your jurisdiction, decision type, and the tool's role, and compare the kit's plans for the record, builder, and review queue. More from Ingenious AI is on the hub.
Related guides
- Colorado AI Act Automated Decision Notices: What to Send
- Adverse-Outcome Explanations in Plain Language
- ADMT Deadline Tracker: Which State Duties Land When
- California CCPA ADMT Requirements: What Your Business Must Do by January 1, 2027
See all ADMT guides for the full set.
Frequently asked questions
What does the free checker ask?
Three questions: jurisdiction (Colorado, California, Connecticut, European Union), decision type (employment, housing, credit/lending, education, or another kind), and how big the AI's role is (decides alone, a substantial factor, or a minor input to a genuine human decision).
Where does the checker's data come from?
Every duty and date was verified against primary or authoritative sources on October 28, 2026 — statute text, official bill pages, the CPPA regulations, EUR-Lex materials — and the sources are listed per jurisdiction on the result page.
What does 'check the list' mean?
For the 'other' decision type, the checker cannot tell whether your decision is on the jurisdiction's statutory list, so it shows the list and the duties that would apply if it is. It is a conditional verdict — a research task, not a refusal.
Is the checker legal advice?
No. Its disclaimer states it gives general information about published duties based on the listed sources, does not judge whether any decision was fair, and that a lawyer should review your setup.
The checker has two datasets — which should I trust?
The public checker (checker_data.py) ships only what could be verified against primary sources, with citations. The account-side rules data was compiled from secondary summaries. Where they differ, build from the checker's verified data.