Prohibited practices
Social scoring, manipulative AI, emotion recognition at work — banned since February 2025. Fines reach €35M or 7% of turnover — SMEs pay the lower of the two. Art. 5
Every document your company needs for the EU AI Act: check in two minutes which obligations apply to you — then close every gap with 24 plain‑language templates in Word and Excel. Written for companies without a legal team, and kept current with every Commission update.
Your result · Provider + Deployer
Applies now
Users must be clearly informed they are interacting with AI.
Machine‑readable marking of synthetic output.
Documented AI‑literacy measures appropriate to your staff and use context.
Ahead — prepare on your own schedule
Deadline moved by the Omnibus — time to prepare, not to panic.
The Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026) rewrote the timeline. Every RegShelf document reflects the amended regulation, with a public changelog whenever the Commission moves.
Social scoring, manipulative AI, emotion recognition at work — banned since February 2025. Fines reach €35M or 7% of turnover — SMEs pay the lower of the two. Art. 5
Providers and deployers must take measures to support staff AI literacy — in force since February 2025, and one of the most immediate obligations for most companies. Art. 4
Chatbot disclosure, deepfake labels, content marking — most duties apply since 2 August 2026; providers of generative systems already on the market before then have until 2 Dec 2026 for Art. 50(2) marking. The part most companies are missing. Art. 50
Documentation, copyright policy, training‑data summary — with real enforcement powers since August 2026. Art. 53 ff.
Hiring, credit, education and similar — moved from Aug 2026 to 2 December 2027. A preparation window, not a reprieve. Annex III
Medical devices, vehicles and other Annex I Section A products — obligations from 2 August 2028; machinery follows its own Section B path since the Omnibus. Annex I
Twenty‑four fill‑in templates in Word and Excel, written in plain language, each mapped to the provisions it helps you address. What applies today, and the full high‑risk preparation set for 2027. No platform, no subscription, no sales call.
Every AI system you provide or use, with your legal role for each — the foundation of everything else.
Baseline
A structured check that nothing you build or buy crosses the red lines.
Art. 5
A practical way to document your Article 4 measures.
Art. 4
A ready‑to‑run training deck with attendance records.
Art. 4
Acceptable use, shadow AI, approval paths for new tools.
Governance
Copy snippets and placement guidance for user‑facing AI.
Art. 50(1)
Implementation guide mapped to the Commission's Code of Practice on marking.
Art. 50(2)
Visible labels for synthetic media and public‑interest text.
Art. 50(4)
Information templates for affected persons, GDPR‑aligned.
Art. 50(3)
What to ask every AI supplier before their risk becomes yours.
Deployer duties
For teams providing general‑purpose models.
Art. 53
Omnibus‑corrected deadlines and a defensible "what applies to us" memo.
All of the above
For AI in hiring, credit, education and essential services — obligations from 2 December 2027. Included in the kit, so you can prepare on your own schedule.
Risk Classification Assessment · Art. 6, Annex III — Technical Documentation · Annex IV — Risk Management Procedure · Art. 9 — Data Governance · Art. 10 — Human Oversight SOP · Art. 14 — Accuracy & Robustness Test Log · Art. 15 — FRIA Template · Art. 27 — Deployer Obligations Checklist · Art. 26 — Instructions for Use · Art. 13 — Post‑Market Monitoring & Incident Kit · Art. 72–73 — Conformity & Registration Walkthrough · Art. 47–49
Plain-English explainers for every part of the Act — and two free downloads to get started.
The whole regulation in plain English — who it covers, what applies when.
Every deadline 2025–2028 after the Digital Omnibus.
The duties in force since August 2026 — chatbots, marking, deepfakes.
The three tiers, and why SMEs pay the lower amount.
What small companies actually have to do — and what they can skip.
Annex III explained: what counts, and what to prepare for 2027.
Every document in the kit, one page each: what the law requires and what the template contains.
A PDF checklist and an Excel inventory — free, no signup.
No. RegShelf provides general regulatory information and working documents — the same starting point a consultant would charge four figures to produce. For decisions about a specific system, especially anything near high‑risk classification, involve qualified counsel. Our templates make that conversation shorter and cheaper.
SMEs, SaaS companies and agencies that build or use AI in the EU and don't have an in‑house legal team. If you run a compliance department with a GRC platform, you are not our customer — and that's fine.
They will. Every pack includes twelve months of updates: when the Commission publishes new guidance or the timeline moves, you get the revised documents and a plain‑language note on what changed. The changelog is public.
Word (.docx) and Excel (.xlsx) — deliberately. Your lawyer can redline them, your team can fill them in, and nothing lives in a platform you have to keep paying for.
Prohibited practices: up to €35 million or 7% of worldwide annual turnover. Most other violations, including Article 50 transparency duties: up to €15 million or 3%. Supplying misleading information: up to €7.5 million or 1%. SMEs and small mid‑caps pay the lower of the percentage and the fixed amount (Article 99(6)).
Providers who place AI systems on the EU market, deployers who use AI under their authority in the EU, and importers and distributors — regardless of where the company is established, as long as the system or its output is used in the EU. Most companies are both provider (AI features they ship) and deployer (AI tools they use).
In force today: the Article 5 prohibitions (since February 2025), Article 4 AI-literacy measures (since February 2025), the Article 50 transparency duties — chatbot disclosure, content marking, deepfake labels (enforceable since 2 August 2026; providers of generative systems already on the market before then have until 2 December 2026 to complete machine-readable marking) — and GPAI model obligations. High-risk obligations under Annex III follow on 2 December 2027.
The Act does not literally prescribe "a policy" — but Article 4 requires measures supporting staff AI literacy, and an internal AI-use policy plus documented training is the accepted, auditable way to evidence them. Both are included in the kit as fill-in templates.
Yes. If the kit doesn't fit your company, write to hello@regshelf.com within 14 days and tell us what was missing — we refund, and your feedback makes the kit better. Since the documents are delivered instantly, checkout includes the standard consent that waives the statutory withdrawal right; this voluntary guarantee replaces it.