← AI Safety Check

LEGAL CHANGES

What changed, and when we last looked

AI Safety Check is maintained on a quarterly review cycle: we check the EU AI Act, its amendments, the GDPR and German trade-secret law, and update the training modules and the policy templates when something changes. This page is the public record of that work — it is our commitment under the Business plan.

LAST REVIEW

2026-09-25

NEXT REVIEW DUE

2026-12-25

TEMPLATE VERSION

2026.3

Current status — what applies now

  • • Article 4 (AI literacy) — in force since 2 February 2025 and rewritten by the AI Omnibus (signed 8 July 2026, Official Journal 24 July 2026, in force 27 July 2026) as an effort duty: companies using AI must take measures supporting AI literacy, taking into account their knowledge, experience, the context of use and the persons affected. No course or certificate is prescribed by law, and Article 4 carries no separate fine of its own. National supervision applies since 2 August 2026.
  • • Article 50 (transparency and marking) — applies from 2 August 2026: chatbots and assistants must disclose they are AI; deepfake-style content depicting real people, places or events must be labelled visibly.
  • • Article 50(2) machine-readable marking — for generative AI systems already on the market before 2 August 2026, the transitional deadline is 2 December 2026. Systems placed on the market from 2 August 2026 must comply when placed on the market.
  • • From 2 December 2026 — new prohibitions apply (non-consensual sexual deepfakes and child sexual abuse material), and national enforcement covers the transparency rules and AI literacy.
  • • High-risk systems — Regulation (EU) 2026/1744 fixes two application dates: high-risk systems under Annex III (Article 6(2) — e.g. employment, education, credit) apply from 2 December 2027, and high-risk systems embedded in regulated products under Annex I (Article 6(1)) from 2 August 2028.

Change log

v2026.3 · 2026-09-25

  • • Added module 10: AI Act marking and transparency duties (Article 50), prohibited practices and evidence.
  • • Unified the Article 50 dates across the site, the training and the audit pack: 2 August 2026 application, 2 December 2026 marking transition for pre-existing generative systems.
  • • Aligned the Article 4 wording with the Digital Omnibus effort duty (no certificate claimed).
  • • Reworked pricing: Training only, Passport S/M/L by team size, Business as an add-on with annual option, Done-for-you service.
  • • Published AGB, AVV/DPA and About pages; added the USt-IdNr. and the legal-changes page with template version stamps.

First published · 2026-09-25

AI Safety Check v2026.3 is the first public release of the programme and of this review cycle. The change log above is the complete record so far; earlier versions were not published.

Legal milestones we work to

  • • 2 Feb 2025 — Article 4 (AI literacy) applies.
  • • 2 Aug 2025 — General-purpose AI model obligations and the prohibition rules apply.
  • • 8 / 24 / 27 Jul 2026 — the AI Omnibus amending the AI Act: signed 8 July, published in the Official Journal 24 July, entered into force 27 July 2026.
  • • 2 Aug 2026 — main body of the Act applies: Article 50 transparency, national enforcement, supervision of AI literacy.
  • • 2 Dec 2026 — transitional deadline for machine-readable marking of AI-generated content from generative systems already on the market; new prohibitions apply.
  • • 2 Dec 2027 / 2 Aug 2028 — high-risk obligations apply: 2 December 2027 for Annex III systems, 2 August 2028 for Annex I systems (Regulation (EU) 2026/1744).

What we check every quarter

  • • EU AI Act: Article 4 (literacy), Article 5 (prohibitions), Article 50 (transparency and marking), high-risk timelines.
  • • Amendments and delegated acts (including Digital Omnibus follow-ups) and Commission guidance.
  • • GDPR developments relevant to employee training records and AI tool vetting.
  • • German law: GeschGehG (trade secrets), works-council and employment aspects of AI policies.
  • • Then we update the modules and the policy templates, bump the version, and publish the change here.

Sources

This page records our reading of the sources above; it is not legal advice and does not replace an individual legal review.