Working paper v0.1 (27 May 2026) of Adopting AI Act Transparency in Three Stages: A Planning Typology for SMEs. This Translational policy essay proposes a three-stage adoption typology — registered disclosure, live attestation, and continuous compliance audit — that lets a small or medium-sized enterprise (SME) locate its current EU AI Act transparency posture, see what artifacts the next stage would add, and decide whether the marginal cost is worth the marginal compliance gain. Each stage is a meaningful compliance posture in itself, not a stub on the way to "real" compliance. The typology is a planning artifact for SMEs and a calibration vocabulary for regulators: it shifts the AI Act enforcement question from "is the deployer fully compliant?" to "which stage has the deployer credibly reached?" The argument's policy implication, developed in §6, is that AI Act enforcement toward SMEs can be calibrated to staged adoption rather than to uniform Article 17 compliance from the high-risk applicability date. Staged enforcement targets the obligations SMEs can actually reach with their current capability and avoids the performative-compliance dynamic that uniform deadlines tend to produce. The framing also helps regulators see which institutional preconditions — particularly the stage-3 auditor profession — still need to be built before the AI Act's full compliance posture is reachable in practice. Companion artefacts. The essay is a policy companion to a separately submitted empirical case study (under double-anonymous review at ACM Digital Government: Research and Practice as manuscript DGOV-2026-0116), which reports integration-cost numbers and architecture detail for a stage-1 deployment in the eAudit building-audit SaaS at eaudit.ee. The two artifacts are deliberately distinct: the DGOV case study reports the numbers; this essay extracts the typology as a standalone policy contribution that does not depend on the case-study data. Status. Submitted to Data per the Cambridge AI policy, AI-assisted writing is disclosed in the manuscript front matter, the author retains full responsibility for the content, and no AI tool is listed as an author. Tyche Institute is a research and education entity, not a trust-service provider. The article does not claim that any actor provides eIDAS trust services, qualified certificates, conformity assessment, or legal compliance services. References to EATF, the EU AI Act, and eIDAS are scholarly and analytic in scope.
Anton Sokolov (Wed,) studied this question.