Learn how StoryChief identifies AI-generated content, what Article 50 requires, and when you remain responsible for adding a visible AI label.

Article 50’s transparency obligations apply from 2 August 2026. They distinguish between two types of disclosure:

A machine-readable mark and a visible disclosure serve different purposes. Having a C2PA credential or an invisible watermark does not automatically remove your responsibility to add a visible label when one is required.

See the European Commission’s Article 50 guidance for the full scope and exceptions.

StoryChief provides technical provenance features, but you control the content, its publication context, and the audience that will see it. You therefore remain responsible for visible disclosures required for your use.

Images and videos generated through StoryChief include:

These signals help compatible tools identify content as AI-generated or manipulated. They are not visible to viewers and are separate from a visible AI label.

Provenance signals are useful indicators, but they do not prove that content is accurate, unedited, legally owned, or presented in the correct context. Metadata may also be removed by editing software, file conversion, or publishing platforms.

When a media file contains a supported C2PA credential, you can review its provenance information in the StoryChief Media Library.

This works for supported media generated through StoryChief as well as files uploaded to StoryChief with an existing C2PA credential.

If StoryChief does not detect a C2PA credential, this does not necessarily mean that the file was created without AI. Its metadata may have been removed, or the tool that generated it may not support C2PA.

Article 50 does not require every piece of AI-assisted content to carry a visible label. The user or organization publishing the content must assess whether the disclosure rules apply.

A visible disclosure may be required in the following situations.

A deepfake is AI-generated or manipulated image, audio, or video content that:

Examples may include a realistic video of a person saying something they never said, an altered photograph of a real event, or realistic footage of an event that never happened.

The assessment depends on the content, message, publication context, and audience expectations. Clearly fictional or stylized content may be treated differently.

A visible disclosure may also be required when AI-generated or materially manipulated text:

Matters of public interest can include politics, public health, public safety, environmental protection, consumer safety, financial developments, science, culture, and other subjects of public debate.

The disclosure may not be required where a qualified person has substantively reviewed the content, a person or organization exercises real editorial control, and that person or organization accepts editorial responsibility for the publication. Spell-checking, formatting, or a superficial review is not sufficient by itself. The European Commission provides more detail in its Article 50 questions and answers.

When a disclosure is required, it must be clear, distinguishable, accessible, and perceivable without requiring a special verification tool. It should appear no later than the viewer’s first exposure to the content.

Use plain, specific language such as:

Choose wording that accurately describes how AI was used. For example, use “Modified with AI” rather than “AI-generated” when authentic content was only partially altered.

As practical guidance:

The EU also provides optional icons for AI-generated and AI-modified content. Using an icon is optional and does not establish compliance by itself. A plain-language label alongside it can make the meaning clearer.

For evidently artistic, fictional, satirical, or creative works, the disclosure may be presented in an appropriate way that does not interfere with enjoyment of the work—for example, in a caption, introduction, or credits. Confirm the appropriate approach for your use case with your legal team.

StoryChief does not currently add machine-readable provenance signals to AI-generated text.

Our text-generation capabilities rely on OpenAI models, which do not currently expose a supported text-provenance signal that StoryChief can carry through to published content. OpenAI has stated that its goal is to expand provenance support to text as standards and tooling mature.

We are monitoring these developments and intend to support suitable text-provenance signals when they become available to integrators.

This technical limitation does not remove your responsibility to add a human-visible disclosure when AI-generated text falls within Article 50’s disclosure requirements. See OpenAI’s current provenance guidance for its latest position.

No. StoryChief automatically includes machine-readable provenance in generated images and videos, but you are responsible for determining whether a visible disclosure is required and adding it to the content you publish.

No. C2PA is machine-readable and may not be visible to your audience. The European Commission explicitly states that deployers cannot rely only on embedded machine-readable marking when a clear disclosure to people is required.

No. Under Article 50, the visible-disclosure requirement for images, audio, and video focuses on content constituting a deepfake. Other laws, platform rules, contractual obligations, or internal policies may impose broader requirements.

No. The Article 50 obligation concerns published AI-generated or manipulated text intended to inform the public about matters of public interest. An exception may apply when the text receives substantive human review or editorial control and a person or organization accepts editorial responsibility.

Confirm:

For authoritative information, consult the European Commission’s guidelines on Article 50 transparency obligations.