Guide · Published November 14, 2026 · Reviewed November 14, 2026 · Ingenious AI team · Not legal advice
GDPR Article 22 vs ADMT Notices: Two Different Regimes, One Careful Pointer
The short answer: The kit's EU checker result ends with a deliberate pointer: the GDPR (Regulation (EU) 2016/679, Article 22) gives people rights over decisions based solely on automated processing — and it is a different regime, outside the checker. This guide exists to keep that pointer accurate: what Article 22 is about at the level the kit states it, how its shape differs from the ADMT notice duties and the AI Act's Article 50, and why the kit covers the US notice-and-record duties while flagging the GDPR as a separate conversation a business with EU decision-making also needs.
Check your duties and dates — free Build your first notice — free
What is Article 22, at the level of this guide? It is the GDPR provision the checker names as giving people rights regarding solely automated decisions. Nothing in the kit's data goes further — no day counts, no procedure, no exceptions are stated — and this guide invents none. Where this guide compares regimes, it compares them on the structural facts the kit's data does state: trigger, geography, and what each is for.
Different triggers: 'solely automated' vs the ADMT role tests
Article 22's stated trigger is a decision based solely on automated processing. The US regimes in the kit deliberately reach further: California covers technology that replaces or substantially replaces human decision-making and expressly fails token review; Connecticut's profiling rights, since July 1, 2026, do not turn on the word 'solely' at all — it was removed. The structural lesson is that 'a human was involved' means different things in different regimes, and a business cannot carry a conclusion from one frame into another.
This is also why the kit keeps the GDPR as a pointer rather than a result: its checker encodes Colorado, California, Connecticut, and the AI Act's Articles 50 and Annex III timeline. A GDPR analysis of a solely-automated EU decision needs the GDPR's own materials and advice, not an extrapolation from a US notice tool.
Different instruments: rights regime vs notice-and-record duties
The ADMT duties the kit implements are notice-and-record duties: tell the person before use, explain afterwards, honour opt-outs and corrections, keep the record. The AI Act's Article 50 is a transparency regime about interaction and content. The GDPR, as the checker frames it, is the data-protection regime under which solely automated decisions carry rights — a broader and older instrument, administered in a different frame, with its own concepts (including the 'solely' line) that the US statutes have consciously drawn differently.
Conflating the instruments produces concrete errors: assuming an Article 50 interaction disclosure discharges a GDPR question, or assuming a Colorado-style notice answers an EU solely-automated-decision right. Each regime's artefacts evidence its own duties. A well-run programme keeps three files — US ADMT, EU AI Act, GDPR — with cross-references, not mergers.
Worked example: Baltic Lending Cooperative
Example — fictional business, real duty structure. Baltic is a fictional lender with a small EU-facing pilot alongside its US operations. Mapping its regimes separately produced this position:
- US ADMT file: Colorado and California duty sets built for the credit tool — assessments, notices, opt-outs, access answers — in the kit's record.
- EU AI Act file: Article 50 position verified (interaction disclosure live on the pilot's surfaces; content marking tested through exports); creditworthiness mapped to Annex III with the December 2, 2027 deployer date diarised.
- GDPR pointer actioned: the pilot's decision flow is flagged for a separate GDPR review because one product variant decides without a human step — the 'solely automated' pattern the pointer names; the review is commissioned as its own engagement, not answered from ADMT materials.
- Shared capture: one per-decision capture discipline serves all three files — reasons, data summaries, dates, and both versions of any re-run — so each regime's analysis draws on the same evidence.
- Register note: the GDPR pointer is recorded on the product's register row with an owner and a review date, so the separate conversation cannot silently lapse.
What the kit does and does not do for EU-facing businesses
The kit's contribution to an EU-facing business is specific: it tracks the EU dates in the same calendar as the US duties (Article 50 live since August 2, 2026; Annex III deployer duties from December 2, 2027), and its checker returns the EU transparency position. Its records cover the Colorado, California, and Connecticut duties. It does not assess GDPR compliance, does not advise on solely automated decisions in the EU, and its documents say plainly that the kit is not legal advice.
Stating that boundary is the guide's main service. Businesses searching for 'automated decision rights EU' can easily land on US ADMT material and misread it as the whole picture; the checker's pointer exists to prevent exactly that. If your operation makes solely automated decisions about people in the EU, the GDPR conversation is a separate engagement — this guide's job is to make sure you know it exists and do not mistake this toolkit for it.
The one workflow insight that does transfer
One discipline transfers across all three regimes without distortion: the per-decision record. Notices given, data used, reasons stated, reviews held, outcomes communicated — dated and producible. Whatever rights a person exercises, under whichever regime, the business answers from the same evidence base. The kit's hash-chained record is built for the US duties, but the capture habit — reasons at decision time, dates on everything, both versions of a re-run — is regime-independent good practice.
Design capture once, at the decision, and let each regime's analysis draw on it. The opposite design — reconstructing per regime when asked — fails identically everywhere: slowly, expensively, and under a clock where one exists. The record method is in How to Keep ADMT Records for Three Years.
Reading the checker result correctly
When the free checker returns its EU result, read its structure: the Article 50 duties as live obligations, the Annex III note where your decision area is listed, the Article 99 enforcement context, and then the GDPR pointer, labelled as a different regime outside the checker. That structure is the honest map of what the tool knows. Use it for what it is — the transparency position and the calendar — and take the pointer as a routing instruction to GDPR materials and advice for the solely-automated question.
The US sections of this guide series make the complementary point from the other side: Connecticut's removal of 'solely' and California's token-review rule mean the US duties often attach where a GDPR-only analysis would stop. Businesses operating across the Atlantic need both maps, drawn separately. Start with the checker for the map the kit can draw, and keep the pointer visible in your register notes so the second map gets commissioned rather than forgotten.
- Keep three files — US ADMT, EU AI Act, GDPR — cross-referenced, never merged
- Do not carry 'a human was involved' conclusions between regimes: the US tests expressly reach beyond 'solely', and the GDPR pointer turns on it
- Treat the checker's GDPR line as a routing instruction: if you make solely automated decisions about people in the EU, commission that analysis separately
- Build one per-decision capture discipline (reasons, data, dates, re-runs) that every regime's analysis can draw on
- Record the pointer on the register row for any EU-facing product, with an owner and a review date
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
- How to Keep ADMT Records for Three Years: What Goes in the File and Why the Chain Matters
- Colorado AI Act Automated Decision Notices: What to Send
- Adverse-Outcome Explanations in Plain Language
- ADMT Deadline Tracker: Which State Duties Land When
See all ADMT guides for the full set.
Frequently asked questions
What does the kit say about GDPR Article 22?
That it gives people rights over decisions based solely on automated processing, that it is a different regime from the AI Act and the US ADMT duties, and that it is outside the checker. The kit states nothing further, and this guide invents nothing further.
Is GDPR compliance part of the ADMT Notice Kit?
No. The kit covers Colorado, California, and Connecticut notice and record duties, and tracks the EU AI Act dates. GDPR analysis is a separate engagement.
Why does 'solely' matter?
Because the US regimes deliberately go further: California covers technology that substantially replaces human decision-making and fails token review; Connecticut removed 'solely' from its profiling rights in 2026. A conclusion under one frame does not transfer to another.
If we comply with Article 50, are we done in the EU?
Article 50 is the transparency layer, already in force. Annex III deployer duties follow on December 2, 2027 for the listed areas, and the GDPR is a separate regime entirely, flagged by the checker's pointer.
What practice transfers across all three regimes?
The per-decision record: reasons and data captured at decision time, every notice and response dated, both versions of any re-run kept, producible on request.