Starting August 2, 2026, a central pillar of the EU AI Act takes effect: companies must make AI systems significantly more transparent to users and market participants. The new rules dramatically increase pressure on providers – non-compliance carries steep penalties.
The essentials
- Article 50 of the EU AI Act establishes concrete transparency obligations effective August 2, 2026
- Chatbots must explicitly disclose their machine nature – users must clearly know whether they're interacting with AI or a human
- Deepfakes (manipulated images, audio, video) require technical markers to identify their artificial origin
- Fines for violations of transparency requirements are provided for in the EU AI Act
- Federal Network Agency (Bundesnetzagentur) becomes the central market surveillance authority under the KI-MIG (Bundesrat, 10 July 2026); BaFin, BSI and BfArM keep their sectors
Four core obligations for companies
The new rules break down into four concrete requirements:
| Obligation | Description |
|---|---|
| Automatic system detection | AI interactions must be identified as such |
| Technical encoding | AI outputs must be uniquely identifiable |
| Deepfake warning | Manipulated content must be explicitly labeled |
| Media content control | AI-generated content in media is subject to control obligations |
The regulation targets disinformation and protects users from deception. Deepfakes – convincingly realistic AI-generated or manipulated recordings – must be immediately recognizable as artificial to viewers.
Who's covered, who's exempt?
Not all AI applications fall under strict labeling rules. Exemptions apply to:
- Tools that merely provide spell-checking functions
- Content clearly recognizable as fictional to the audience with no deceptive intent
This leaves room for legitimate uses – such as obvious satire or pure writing aids. However, providers bear the burden of proving their solution qualifies for an exemption.
Who supervises this in Germany
Enforcement in Germany is settled by the AI Market Surveillance and Innovation Promotion Act (KI-MIG): adopted by cabinet on 10 February 2026, passed by the Bundestag on 11 June 2026, approved by the Bundesrat on 10 July 2026 — just under three weeks before the obligations start to apply.
The Federal Network Agency (Bundesnetzagentur) becomes the central market surveillance authority. It serves as the central point of contact and complaints, and is building a dedicated AI market surveillance chamber alongside a coordination and competence centre.
It is not the only authority, though. BaFin, BSI and BfArM retain supervision in their sectors, so which regulator applies to your system depends on where you deploy it, not only on what it does.
What violations cost
Companies ignoring the new requirements face severe penalties. The EU AI Act provides for fines for transparency violations. This is serious money – potentially existential for mid-sized firms.
The message is clear: implementation is mandatory, not optional.
What this means for your business
If you deploy or provide AI systems – whether chatbots, image generators, or content tools – you should act now. The deadline until August 2026 is tighter than it sounds: you need time for technical adjustments, internal training, and documentation. Critical step: work with your legal team on two questions — which of your systems fall under the labeling requirements and which are exempt, and which of the four authorities is the competent one for you.
Sources
- Deutscher Bundestag — Ja zur Durchführung der Verordnung über künstliche Intelligenz
- Bundesregierung — Umsetzung der KI-Verordnung
- Bundesnetzagentur — Marktüberwachung KI
- AD HOC NEWS
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
