Rechtsstand

THE ARTICLE THAT GOVERNS EVERY OTHER DATE, and it was missing from this index until 2026-09-03. Verified verbatim against buzer.de/113_KI-VO.htm. Art. 113(3)(a): Chapters I and II apply from 2025-02-02, EXCEPT Art. 5(1) subpara 1 points (ba) and (bb) and Art. 5(1a)/(1b), which apply from 2026-12-02. Art. 113(3)(c): Chapter III Sections 1, 2 and 3 — which contain Art. 13 (instructions for use) and Art. 26 (deployer obligations for HIGH-RISK systems) — apply from 2027-12-02 for Annex III high-risk systems and from 2028-08-02 for Annex I. CONSEQUENCE, and it is the one most German content misses: legal databases print 'Geltung ab 02.08.2026' at the head of Art. 26, but that heading is expressly subject to Art. 113. For a deployer with no Annex III purpose the Art. 26 duties are not merely inapplicable by subject matter — they have not started.

Obligation softened from 'ensure' a level of AI literacy to 'take measures to support' its development; the Omnibus states expressly that it does not require providers or deployers to guarantee any specific level of AI literacy of any individual.

Definitions of the Regulation. Cited here because the provider/deployer distinction in Art 50 is only decidable through them.

Art 3 Nr. 3 defines the PROVIDER: a natural or legal person that develops an AI system or has one developed and places it on the market or puts it into service under its own name or trademark. Art 50(1) and 50(2) bind this actor.

A deployer is a natural or legal person using an AI system under its authority, except for personal non-professional activity. Load-bearing for this product: the pack must never attribute a provider duty (Art 50(1)/(2)) to a deployer, and Art 25 can convert a deployer INTO a provider — which is why the pack states the addressee of each paragraph instead of deciding the customer's own role.

NEW Article. Creates a legal basis for processing special-category personal data to detect and correct bias, subject to six cumulative conditions including strict necessity. Extends to all providers AND deployers, not only providers of high-risk systems.

Periodic penalties capped at 5% of average daily income or worldwide annual turnover in the preceding financial year, per day.

Code AIH 0401 — 'emerging AI technologies not covered by other codes, including agentic AI'. First appearance of 'agentic AI' in binding EU law. No definition, no obligations, no classification criteria: competence-designation scaffolding only.

Art 50(2) synthetic-content marking: systems placed on the market before 2 August 2026 must comply by 2 December 2026, in a machine-readable format.

AI Office gains exclusive enforcement competence over GPAI systems where model and system come from the same provider or undertaking, and over systems integrated into VLOPs/VLOSEs. Deployers forming part of the same undertaking as the provider fall under AI Office exclusive supervision.

Art 50(3) binds the BETREIBER: deployers of an emotion-recognition or biometric-categorisation system must inform the natural persons exposed to it. One of only two Art 50 paragraphs that bind the deployer rather than the provider. In force since 2 August 2026.

Art 50(4) binds the BETREIBER, in two cases: (1) deployers generating or manipulating deep-fake image/audio/video content must disclose that it is artificially generated; (2) deployers publishing AI-generated TEXT to inform the public on matters of public interest must disclose it — with an express carve-out where the content underwent human review or editorial control and a natural or legal person holds editorial responsibility. The live deployer transparency duty this product exists to evidence.