AI video disclosure:
the agency handoff checklist.

The production team may create dozens of AI-assisted variants. The client, channel, or public only sees one final master. That is the moment to preserve what was actually approved—not what the team intended to review earlier in the process.

1. Treat the final cut as a decision

AI production makes it easy for the final version to drift: a late colour change, a shortened social cut, a different voice-over, or a new export can change what was reviewed. A useful handoff identifies the exact file or final URL, the person who approved it, and the publication context.

This is ordinary production hygiene as much as it is a transparency workflow. It helps a producer answer a client question later without reconstructing a campaign from messages, folders and memory.

2. Separate production facts from legal scope

Article 50 contains specific transparency obligations for particular systems and content. The European Commission’s guidance describes deployer disclosure for content that constitutes a deepfake and for certain AI-generated or manipulated text published to inform the public on matters of public interest. It also distinguishes provider machine-readable marking from a deployer’s audience-facing disclosure.

That does not mean every AI-assisted video receives the same answer. Record the facts first: what was generated or manipulated, whether the final media could appear authentic or truthful, the intended audience and channel, and who is responsible for publication. Use qualified legal advice for a decision that depends on a specific campaign or jurisdiction.

3. Build the final-master handoff record

Keep the record short enough to be used every time. The point is to make the final step repeatable, not to create a second production system.

• Identify the final asset: campaign, client, format, channel, planned publication date and final file fingerprint or final URL.
• Describe the AI contribution: generated footage, manipulated footage, assistive edit, voice, image, sound, or mixed work.
• Record the review: reviewer, date, what changed after review, and the person or organisation responsible for publication.
• Keep the source context: tool/model where known, provider marking/credential status, and any supporting approval link.

4. Preserve the disclosure decision

When a team decides that an audience-facing disclosure is needed, record the exact wording and where it appears: title card, caption, platform label, description, spoken notice, or another placement. A label that is only proposed in a briefing is not evidence that it appeared with the final master.

Where the team decides no such disclosure is required, record the reason in plain language and the evidence reviewed. That is useful for future consistency and for explaining the decision to a client. It is not a legal determination.

5. The seven-minute final-master handoff

1. 01
   Freeze the version.

   Save the final file or URL and generate a fingerprint for the exact export.

2. 02
   Name the channel.

   Record where and when the audience will encounter the asset.

3. 03
   Describe AI’s role.

   State the material contribution, not simply that a tool appeared somewhere in production.

4. 04
   Record the real review.

   Name who reviewed the final version and any consequential changes.

5. 05
   Check the disclosure path.

   Record the relevant facts and route any uncertain legal question to the responsible adviser.

6. 06
   Copy the exact wording and placement.

   Where used, preserve the audience-facing notice and the place it appears.

7. 07
   Export the record.

   Give the client team a portable handoff while the facts are still fresh.

Primary sources and related reading

European Commission: Guidelines on transparency obligations for providers and deployers of certain AI systems.

Regulation (EU) 2024/1689, Article 50. Read the Article 50 agency checklist for the broader workflow.