What If I Disagree With My Classification?
The result is evidence-based — and revisable.
Classifications are deterministic: they follow from your answers and the Act's rules. If the result looks wrong, one of three things is happening.
1. An answer doesn't reflect the system
Most disagreements trace back to one question answered from the wrong perspective — usually "does the output materially influence a decision about a person?". Re-read your answers with the engineer and product owner together, then re-run the assessment. Classifications are free to re-run on every plan tier's included quota.
2. Your case sits in a genuine grey zone
Some systems legitimately straddle Annex III boundaries — the Act is new and case law doesn't exist yet. In that situation:
- Run the assessment both ways and keep both rationales.
- Document why you adopted the lower classification, including the Art. 6(3) reasoning.
- Take that package to legal counsel. A documented, defensible position is what matters if a regulator asks.
3. You found a real misclassification
If you believe the rule mapping itself is wrong — the questionnaire put you in the wrong Annex III category — contact support with the system description and your reasoning. We review every report; the rule base is versioned and corrections ship quickly.
Remember: Guardia AI provides an indicative assessment, not a legal determination. The final call on a borderline system belongs with qualified counsel.