Sources & verification
Every claim, traced and dated
We don't ask you to trust us — we show our work. Each key regulatory claim on this site is traced to a first-party source and dated. Where we rely on analyst reporting rather than a primary text, we label it an estimate.
Last verified: 31 August 2026 · Re-verification routine: quarterly, and on any major regulatory event.
| Claim as used on the site | Type | Primary / cited source | Verified |
|---|---|---|---|
| Article 50 transparency obligations are in force; enforcement began 2 Aug 2026 | Primary | EU AI Act Art. 50; European Commission AI Office communications | 2026-08-31 |
| Transparency grace period for existing generative systems cut from 6→3 months → hard deadline 2 Dec 2026 | Primary | EU AI Act Art. 50; Commission AI Office; transparency Code of Practice | 2026-08-31 |
| High-risk deadlines: standalone HRAIS 2 Dec 2027; embedded in regulated products 2 Aug 2028; sandboxes 2 Aug 2027 | Primary | EU AI Act; Council/Parliament final approval 29 Jun 2026 | 2026-08-31 |
| Prohibited practices (incl. certain deepfake/nudifier uses) — in force since Feb 2025; nudifier/CSAM prohibitions hold 2 Dec 2026 | Primary | EU AI Act Art. 5 | 2026-08-31 |
| Penalties up to €35M / 7% of global turnover; €15M / 3%; €7.5M / 1% | Primary | EU AI Act Art. 99 | 2026-08-31 |
| ISO/IEC 42001 is the AI management system standard; Annex A control areas A.2–A.10; certification by accredited bodies | Primary | ISO/IEC 42001:2023 | 2026-08-31 |
| ISO 42001 increasingly named in RFPs as a procurement baseline; SOC 2/PCI seen as insufficient for AI | Analyst-reported | fintech.global (18 Aug 2026) | 2026-08-31 |
| SME / small-mid-cap simplification may extend to ~750 staff / ~€150M | Estimate | Analyst reporting; not itemised in the 29 Jun 2026 Board text reviewed | 2026-08-31 |
| Indicative ISO 42001 first-year cost ~US$37.5–85k (SMB, all-in) | Estimate | Industry/analyst reporting (varies by scope & body) | 2026-08-31 |
How to read this. Primary means the claim is grounded in the regulation or standard text itself. Analyst-reported / Estimate means we rely on secondary reporting or an indicative figure — treat these as directional until confirmed in a primary text. None of this is legal advice; for binding interpretation in your case, consult counsel.
Spotted something out of date? Tell us at bigjay11@gmail.com — accuracy is the product.
Our verification methodology & QA routine
Accuracy isn't a one-off. Here is the process that keeps every regulatory statement on this site — and inside every document the builder generates — trustworthy.
How each claim is verified
- Trace to a first-party text (regulation article or standard clause) wherever possible.
- Where only secondary reporting exists, label it Analyst-reported / Estimate and never present it as settled law.
- Record the exact article/clause and the date checked.
- Cross-check the same fact everywhere it appears (site pages, tools, generated documents) so nothing contradicts.
Re-verification schedule
- Quarterly review of every claim in the table above.
- Event-driven review on any major regulatory move (new delegated act, guidance, deadline change) — surfaced in Watch.
- Last verified: 31 Aug 2026 · Next scheduled review: 30 Nov 2026.
- Every generated document carries its own “references current as of” note so your evidence is dated too.
Internal consistency check (last run 1 Sep 2026): the near-term deadline (transparency 2 Dec 2026), the high-risk dates (2 Dec 2027 / 2 Aug 2028), enforcement start (2 Aug 2026) and penalty tiers (€35M/7% · €15M/3% · €7.5M/1%) were confirmed to match across every page, tool and generated document — no contradictions. The AI Act obligations finder, transparency kit and high-risk pack draw from this same verified set.
See these dates applied to your own systems:
Find what applies to you — free →