Legal
AI content labeling & transparency
Last updated July 31, 2026
Everything produced on FlowPixel AI is artificially generated. We mark it that way, in machine-readable form, and we ask you to keep those markings intact when you publish.
What we mark, and how
- Provider provenance is preserved. Model providers embed watermarking and C2PA-style provenance data in their output. We deliver the provider's file as we received it and never re-encode it to strip those markings.
- Container metadata. Exports carry metadata fields identifying the content as synthetic, naming the generating model and the generation timestamp.
- A disclosure receipt. Each render has a downloadable receipt recording the model, settings, generation time, and an explicit AI-generated declaration you can keep for your records or hand to a platform.
- In-product labeling. Videos in your library and every preview are visibly badged as AI-generated.
Why this exists
Article 50 of the EU AI Act requires providers of generative AI systems to mark synthetic audio, image, and video output in a machine-readable format that is detectable as artificially generated, with those obligations applying from August 2026. Deployers who publish deep-fake style content have their own disclosure duty. We build the marking in as a default rather than as an option so that using this service does not create a compliance gap for you.
Similar duties exist elsewhere: several US states require disclosure on synthetic political advertising, and the major video platforms require creators to flag realistic synthetic content when uploading.
Your obligations when you publish
- Do not remove, alter, or obscure watermarks, metadata, or the disclosure receipt.
- Use the platform's own AI-content toggle when you upload to YouTube, TikTok, Instagram, or similar. Our metadata does not switch that on for you.
- Add a visible disclosure where the content could reasonably be mistaken for real footage.
- Never present generated video as authentic recordings of real events or real people.
Limitations we will be honest about
Watermarking is not tamper-proof. A determined re-encode can degrade or remove it, and detection tools are imperfect. Machine-readable marking makes provenance discoverable in ordinary handling; it is not a guarantee against deliberate stripping. Removing our labeling is a breach of the Terms of Service and grounds for termination.
Reporting mislabeled content
If you find content generated here that is circulating without its labeling, or that is being passed off as real footage, report it through the reporting form.