Last updated by Wing Chan on August 20, 2026.
You can read our research post on the EU AI Act, published on August 7, 2026.
This guide explains how to think about Riverflow outputs under the EU AI Act’s transparency rules, especially where customers are creating product imagery, campaign creative, ecommerce visuals, or UGC-style content.
This is general guidance, not legal advice. If you are publishing high-risk, regulated, political, medical, financial, or testimonial-style content, you should take your own legal advice.
Summary
Most Riverflow outputs are best understood as AI-generated or AI-edited creative, not automatically as “deepfakes”.
The key question is not simply whether AI was used. It is whether someone viewing the content could reasonably mistake it for real-world proof of a real person, event, place, endorsement, product result, or factual claim.
If there is a real risk of that kind of confusion, add a clear disclosure.
Key Dates To Be Aware Of
The main date customers need to know is 2 August 2026. From that date, Article 50 of the EU AI Act’s transparency rules applies.
From 2 August 2026:
If you publish AI-generated or AI-edited image, audio, or video content that qualifies as a deepfake, you must disclose that it was artificially generated or manipulated.
If you publish AI-generated or AI-edited text to inform the public on a matter of public interest, you may also need to disclose that AI was used, unless the content has been reviewed by a person and someone takes editorial responsibility for it.
Content generated before 2 August 2026 does not need to be labelled retroactively under the Commission’s guidance, although labelling is encouraged where it helps avoid confusion.
You may also see a 2 December 2026 date mentioned in EU materials. For Riverflow customers, this is usually not the main action date. It mainly relates to a limited transition period for certain AI systems already on the market before 2 August 2026 to meet machine-readable marking rules, plus new prohibitions around certain illegal synthetic content. Riverflow does not permit that type of illegal content in any case.
What Is A Deepfake?
Under the EU AI Act, a deepfake is AI-generated or manipulated image, audio, or video content that resembles existing people, objects, places, entities, or events and would falsely appear to a person to be authentic or truthful.
The European Commission guidance says three things matter:
Does the content resemble a real or plausible person, object, place, entity, or event?
Is the resemblance strong enough to look realistic?
Could the content falsely appear authentic or truthful in the context where it is published?
If the answer is yes, treat it as needing disclosure.
Typical Riverflow Use Cases
For normal Riverflow use cases, such as product shots, product-in-scene images, packshot adaptation, ecommerce visuals, campaign creative, and background generation, the output is usually best understood as AI-generated or AI-edited commercial creative.
That does not mean you can use it to mislead people.
You should not use Riverflow outputs to materially misrepresent:
what a product looks like
what a product can do
the size, quality, ingredients, certification, or performance of a product
whether a real customer, employee, founder, influencer, or celebrity used or endorsed the product
whether an event, review, test, result, or before/after image actually happened
A simple rule: AI creative is fine. Fake evidence is not.
Higher-Risk Content
Be especially careful with content that looks like real-world proof, including:
synthetic customer testimonials
UGC-style videos or images showing someone apparently using a product
fake reviews, star ratings, or customer quotes
realistic founder, employee, influencer, or celebrity likenesses
before/after claims
regulated product claims, including health, medical, financial, safety, or sustainability claims
content that looks like documentary evidence of a real event
synthetic screenshots, receipts, certifications, lab results, or official-looking documents
For these formats, disclosure alone may not be enough. The underlying message must also be true, authorised, and evidence-backed.
UGC-Style Content And Testimonials
If you create UGC-style content in Riverflow, it should be grounded in something real.
Good practice:
Use real customer feedback, reviews, or approved campaign claims as the basis.
Keep a record of the source feedback or approval.
Do not invent a customer experience.
Do not create a synthetic person who appears to be a real customer unless that is clearly disclosed.
Do not imply that a real person said, did, used, tested, or endorsed something unless they actually did and you have permission to use it.
Example disclosure:
“AI-generated visual based on real customer feedback. Person shown is synthetic.”
If you are using a real person’s name, face, voice, quote, or likeness, get explicit permission.
How To Label AI-Generated Or AI-Edited Content
If content could be mistaken for real-world proof, label it clearly at the point someone first sees it.
Good labels include:
AI-generated image
AI-edited image
AI-generated product visual
AI-generated scene. Not an actual customer photo.
AI-generated visual based on real customer feedback. Person shown is synthetic.
For video or audio, put the disclosure somewhere people will notice it, such as in the video overlay, caption, title, description, or audio introduction. If the asset is likely to be downloaded, reused, or shared outside its original context, consider embedding the disclosure into the asset itself.
Do not rely only on hidden metadata, watermarks, or provenance information. The Commission guidance says deepfake disclosures should be understandable and perceivable without requiring special tools.
The EU has also published optional AI-labelling icons. You do not have to use those icons, but any disclosure still needs to be clear and distinguishable.
Practical Checklist Before Publishing
Before publishing a Riverflow output, ask:
Is this just creative product imagery, or could it be seen as proof of something real?
Does it show or imply a real customer, employee, influencer, celebrity, event, location, result, or endorsement?
Would a viewer be misled if they did not know AI was used?
Are all product claims accurate and evidence-backed?
Do we have permission for any real person, quote, likeness, voice, or endorsement?
Should we add a visible AI disclosure?
If in doubt, label it.
Sources
This guide is based on official EU materials current as of 4 August 2026: