How does PrismPoster align to EU AI Act Article 50?
PrismPoster aligns to EU AI Act Article 50 with two signals on every export: a visible label and an always-on machine-readable mark. The visible "AI-generated with PrismPoster" label maps to the Article 50(4) transparency and deepfake disclosure duty, and the embedded machine mark maps to the Article 50(2) requirement that AI output carry a machine-readable signal. Both are mandatory on every plan, including Free. This alignment is metadata-based plus a visible label, not a cryptographic C2PA Content Credentials signature. PrismLabs OÜ is an Estonian company, so EU transparency rules sit at the centre of how exports are built.
What does Article 50 require?
Article 50 of the EU AI Act sets transparency duties for AI-generated content. Two parts matter here. Article 50(2) requires that AI-generated output be marked in a machine-readable format so systems can detect it. Article 50(4) requires that deepfakes and certain AI-generated or manipulated content be disclosed in a way people can understand. PrismPoster addresses both rather than just one.
How does PrismPoster meet 50(2)?
PrismPoster meets the machine-marking duty with the always-on provenance metadata embedded in every file. The mark is written into IPTC, XMP, and EXIF for images, FFmpeg metadata for video and audio, ID3 for audio, and R2 object headers for stored objects. This machine mark is never gated by plan tier, so it's present on all AI output regardless of what you pay.
How does PrismPoster meet 50(4)?
PrismPoster meets the human-facing disclosure duty with the visible label rendered into every export. Fully generated work reads "AI-generated with PrismPoster"; work edited with Magic Edit reads "AI-enhanced." See generated vs enhanced for the distinction and why content is labelled for how the label appears.
Why metadata instead of C2PA Content Credentials?
PrismPoster uses embedded metadata and a visible label, not a cryptographic C2PA signature. C2PA Content Credentials cryptographically bind provenance to a file so it can be verified and tamper-detected; that is not implemented in PrismPoster today. The metadata approach uses standard, widely readable fields and a visible mark on the rendered file. The trade-off is that embedded metadata can be stripped by some platforms on upload, whereas a cryptographic signature is designed to be verifiable. The visible label remains part of the export regardless.
Does this make my content fully compliant for my use case?
PrismPoster supplies the labelling and provenance signals on export; how you use, disclose, and distribute your content is your responsibility. Article 50 obligations can also fall on deployers and users depending on context. For your own legal position, review the terms, the acceptable use policy, and seek your own advice where needed.
Frequently Asked Questions
Does PrismPoster use C2PA Content Credentials?
No. Provenance is metadata-based plus a visible label. Cryptographic C2PA signing is not implemented. See provenance metadata for the fields PrismPoster writes.
Which part of Article 50 does the visible label cover?
The visible label maps to Article 50(4), the human-facing transparency and deepfake disclosure duty. The embedded machine mark maps to Article 50(2), the machine-readable marking duty.
Is the Article 50 alignment available on the Free plan?
Yes. Both the visible label and the machine-readable mark are mandatory on every export and every plan, including Free. The machine mark is never gated by tier.
Where can I read more about PrismPoster's compliance posture?
See the security page and the privacy page. For specific questions, email support@prismposter.com.