APRIL 2026 · INDEPENDENT ANALYSIS
What we found screening 50 European AI companies against the 6 regulatory gates of the EU AI Act. 4 months before Annex III becomes applicable.

50 companies · 4 sectors · 15+ countries · Screened April 10–12, 2026
11 pages · A4 · 753 KB
HEADLINE
72% of the 50 companies screened trigger high-risk classification under Art. 6(1) safety components or Art. 6(2) + Annex III use cases. 96% have no public AI Act position. 0 of the 14 companies already holding sectoral certifications (CE MDR, CE-IVD, ACPR banking licences) have publicly mapped the AI Act layer on top.
44% deploy AI systems with direct end-user interaction — chatbots, voice assistants, medical scribes, public-space video analysis — without documented Art. 50 transparency compliance. 34% integrate general-purpose AI models (LLM or foundation models, external or proprietary) within the meaning of Art. 3(63), with no documented obligations under Art. 53 or Art. 50 + Art. 26.
The compliance gap is not about knowledge — the AI Act is covered in trade press and flagged by industry bodies across every sector in this sample. The gap is classification: these companies have not produced a dated, article-mapped document answering the question "what is our regulatory position?". Until that document exists, every other compliance effort is built without a foundation.
METHODOLOGY NOTE
The 50 companies were deliberately drawn from four sectors with high regulatory exposure (HealthTech, FinTech, HRTech/EdTech, B2B SaaS/Industrial). The sample is dominated by France (~70%) and is a convenience sample, not a statistical panorama of European AI. The 74% high-risk figure reflects this sectoral selection: in HealthTech and regulated finance, high-risk classification is largely structural under Annex III rather than a discretionary finding.
GPAI is counted per Art. 3(63): LLM/foundation models (external or proprietary), excluding narrow task-specific ML such as credit scoring, medical imaging classifiers, or recommendation engines. All percentages are computed on /50. Five entries returned INCERTAIN on at least one gate and are counted in the base but not in the high-risk total.
All profiles in the report are anonymized. Every evaluation is based exclusively on publicly available information — company websites, product pages, CE mark disclosures, regulatory filings, LinkedIn profiles, press releases. No company was contacted.
What this report is — and is not
This is an independent analysis based on publicly available information, not a legal opinion or a formal audit of any company. Classifications are estimates — only a personalized screening can confirm definitive regulatory status. Sprinkling Act operates as a pre-conformity advisory firm in this market; no commercial relationship with any company listed existed at the time of this analysis.
PREVIEW
The first three pages are shown below — cover, executive summary, findings 1–6, and the sector breakdown. The rest of the report (cases, screening tables, enforcement timeline, full methodology) is in the PDF.



Five anonymized case studies, the full 50-company screening tables, the enforcement timeline, penalty structures, and the complete methodology — all in the downloadable PDF.
Or keep reading the closing page below

INDEXED RESEARCH
This report is indexed as an official Sprinkling Act research document. You can verify it at any time from the public AI Act registry by looking up the badge code below.
Badge code
SA-REPORT-2026-04
APPLY IT TO YOUR OWN SYSTEM
The same 6-gate framework used in this report is available as a free 9-question diagnostic. 60 seconds. Zero data collected.
Run the free diagnosticUSAGE
This report is free to download, share, and reference with attribution. A permanent URL is provided above. If you cite it in a compliance note, a board deck, or a regulatory discussion, we simply ask that you link back to this page so readers can reach the full document.