AI transparency statement
This statement implements the transparency obligations of Article 50 of Regulation (EU) 2024/1689 (AI Act), which apply from 2 August 2026. The German version is the binding one.
1. Provider and scope
The provider of the service is Fade Media GmbH, Beispielweg 1, 10115 Berlin, email: info@postomnia.com. This statement covers the AI features inside PostOmnia and AI-assisted content on this website.
2. Which AI systems we use
- Anthropic Claude (Claude API, Anthropic PBC) – produces text suggestions in the composer: caption ideas, hashtag suggestions, platform-specific rewrites and the draft of a YouTube thumbnail as a graphic. The model in use is configured on the server and may change.
- Google Gemini (Gemini API, Google Ireland Ltd. / Google LLC) – generates and edits images in the Creative Studio (visuals, carousels, campaign packs).
Not every feature described as “automatic” is an AI system: templates, quality checks, timing suggestions, inbox reply templates and the local fallback image used when no image provider is configured are rule-based. They are deliberately not labelled as AI-generated, because a false label is just as misleading as a missing one.
3. Notice of interaction with an AI system (Art. 50(1))
Wherever PostOmnia shows the output of an AI system – in the composer, in the Creative Studio and through the interfaces for AI assistants (MCP/REST) – it states that the output was produced by an AI system. Results are suggestions; nothing is published without your involvement.
4. Machine-readable marking of generated content (Art. 50(2))
Images and graphics produced by an AI model automatically receive a machine-readable provenance
marker following the IPTC standard (DigitalSourceType = trainedAlgorithmicMedia) plus a
readable credit line naming provider and model. In PNG files it is embedded as an XMP packet, in SVG
graphics inside a <metadata> block.
Two limits stated plainly: we do not sign provenance cryptographically (no C2PA), and many social platforms strip metadata when they re-encode an uploaded image. The marking therefore survives download and file sharing, but not necessarily processing by a platform.
In addition, PostOmnia records for every post whether the text was generated, by which provider and model, and whether it was edited afterwards. That information is visible in the composer, the calendar, the post list and the client approval portal.
5. Visible labelling of published posts
Every workspace can decide under Settings → AI labelling that a visible note is appended automatically to each published post with AI provenance (wording configurable, default: “Dieser Beitrag wurde mit KI-Unterstützung erstellt.”). The default is off, because a visible label is not required by law in every case and changes every published text. Whether an obligation applies in a given case is for the workspace to judge – see section 6.
6. Obligations of users (Art. 50(4))
Anyone publishing through PostOmnia is a deployer within the meaning of the AI Act for the published content and must disclose that content is artificially generated or manipulated, in particular
- for image, audio or video content that convincingly reproduces real people, places or events (deepfakes), and
- for AI-generated text published to inform the public on matters of public interest.
Independently of that, the labelling rules of the respective platform and the duty to label advertising apply. For ordinary marketing posts that are reviewed editorially there is no general legal duty to mark every AI-assisted text as such.
7. Human oversight
AI output is never published directly. It lands in the composer or as a draft and is released by a human; agencies can additionally use the client approval portal, which flags AI-assisted posts. The only feature that produces text without another click is evergreen recycling: if it is enabled for a post, PostOmnia rephrases that post's text at the configured time and creates a scheduled post from it – visibly flagged and editable or deletable at any time.
8. Rights in AI results
Rights in the generated results lie with the respective workspace as far as legally possible. Purely machine-generated content is regularly not a personal intellectual creation within the meaning of § 2(2) UrhG and therefore not protected by copyright; it is not exclusive. That does not replace a check for third-party rights – details are set out in § 7a of the Terms of Service.
9. No training on your content
Neither PostOmnia nor the providers we use employ the content transmitted through the service to train AI models under the applicable API terms. Which data is sent to which provider is described in section 4a of the Privacy Policy.
10. AI literacy (Art. 4)
Article 4 of the AI Act obliges providers and deployers to ensure a sufficient level of AI literacy among the people who operate AI systems on their behalf. The basics – how the systems work, typical failure modes such as fabricated facts, copyright and personality rights, and what to do in doubtful cases – are summarised in our AI compliance documentation.
11. Contact and changes
Questions about this statement: info@postomnia.com. We update it when the systems we use or the legal situation change.
Last updated: 2026-08-02