Use this KVKK checklist before publishing an AI video
Review faces and voices before publication. Check the vendors and their retention arrangements, along with the required approvals.
Get the KVKK and personal-data review going before anyone uploads footage for an AI video. The same applies before a voice sample is recorded or a synthetic avatar is published. AI can take some of the load in the content workflow. The team still has to decide how identifiable people, and their data, will be handled. It also has to work out which approvals are needed and who carries the responsibility.
We work on technical and education communications. We also handle communications in healthcare and finance, and for the public sector. Communications and marketing teams can pick up this checklist when they plan production with the client's legal or privacy advisers. Project teams can use it the same way. It comes into its own when a video involves employees or students. The same holds for customers and patients, and for executives or representatives. Our broader AI video compliance resources are the place to start. Then go through the questions below to spot what's missing from the workflow. Working through them also puts the review on record.
Identify every face, voice, and personal detail
What in the material could give away who somebody is? List every one of those elements. Faces and voices are on that list. So are names and job titles, along with contact details and locations. ID cards count too, as do screen recordings and whatever sits in the background. A short interview clip may hold several of these elements.
For each person, note why you need the material and where it came from. Then record who can get at it and where the finished video will appear. Count in paid media and websites. Social channels and internal platforms go on the list too, along with event screens and presentations. Look at the context as well. Approval given for an internal meeting may not cover a public campaign, and the campaign may need an approval of its own.
Review client-supplied material with the same care. Receiving a file from the client doesn't establish that it is ready for AI processing. Ask which use has been approved and who confirmed it. Have those answers before the project starts.
Set clear boundaries for synthetic avatars and cloned voices
Synthetic avatars and cloned voices get an approval step to themselves. Confirm whether a real person's image or voice will go in as training material. It might serve as reference material instead, or as the basis for generated output. Write down the script you intend to use and the channels. Note the campaign period, and the person or team who can approve changes.
Who gave the go-ahead for the use? Write down which channels and what period that authorization covers. Record whether the asset can be reused or changed for a new script. Note the same for a new format or a new campaign. File the written consent or authorization alongside the project approval materials. Keep it somewhere the team can lay hands on it while they produce and review the work.
Corporate communications can carry added risk when an avatar seems to speak for someone. That someone could be an executive or an employee. It could be a customer, a student or an expert. Could a viewer mistake the generated message for something that person actually said? Check for that. The brief should say what the output is. It may be fictional, or an approved spokesperson asset. It may instead be a recreation of a known individual.
Keep source recordings apart from the general project assets. Only people who need them for production should have access. If the campaign changes direction, take another look at the approval. The approval for the original asset may not cover every later use.
Apply data minimization to AI video production
Trim the material down to what the approved video requires. Files may be headed for a platform that edits, generates or transcribes, or for an avatar platform. Before they go up, strip out any identifiers that play no part in the work. Go through the scripts and the footage. Check the voice recordings and screen captures as well, and the production notes.
A product demonstration may call for a presenter and some interface visuals. There may be no need for a customer database to show up on screen, though. The same may go for an internal email thread or employees' contact details. Reframe the shot, or swap in sample screens designed for the job. Another route is to prepare a clean script. Choices like these cut down how much personal data the production handles.
Decide how the files will be stored before the work starts. Record where the source files and generated outputs will live. Do the same for prompts and transcripts, and for review comments and exports. Our privacy information points out that people may ask for information about their data. They may also ask for inaccurate data to be corrected, or for deletion once processing is no longer required. Whoever deals with one of those requests needs clear project records to work from.
Review vendors, access, retention, and deletion
An AI video can pass through several tools on its way out. Storage and transcription may both be involved. Editing could be in the mix, and image or voice generation too. Rendering and publishing platforms may play a part as well. List every vendor the project uses. Record what data goes to each service, and who the project owner is. Note what access internal and external contributors have been given.
Before a tool is put to use, ask where it will process the data. Does the vendor use source footage or voice samples for model training? What about prompts or outputs? Which subcontractors can the data be passed to? Read through the data-processing terms. Check any international-transfer terms that apply as well.
Look, too, at what the tool needs in order to do its job, and what it holds on to afterwards. Can collaborators download files? Do generated assets stay in the workspace? Who can clear them out when the project closes?
Let the production plan decide who gets access. Each contributor gets only the files and permissions their work calls for. When that role comes to an end, take the access away. Make a retention decision for the source assets and the generated materials. Then put someone in charge of confirming deletion or archiving when the project closes. Keep a clear record, so the client can see where materials were handled during production.
Escalate sensitive-sector approvals early
Healthcare and education projects need a clear approval route from the start. Finance and public-sector projects do too. The content may involve vulnerable audiences or confidential information, as well as regulated messaging or legal duties. Arrange the approval before the final edit, when deadlines can put pressure on the team to publish.
On education projects, check whether the footage shows students or parents. Look out for classrooms as well, and for names or records. We produced an interview-led campaign for an educational institution, with parents and students in front of the camera. It backed the school's enrollment communications on its website and in search engines. It also ran on social media channels and in outdoor advertising. The example doesn't prove legal compliance. It shows why participant approval and the publication channels need to be sorted out early in planning.
In healthcare and finance, and in public-sector work, identify the client stakeholders who must review the scripts and source materials. They need to do so before production begins. Ask whether anyone shown in the video, or any information in it, needs approval from someone outside the main marketing team. If the answer isn't clear, put the asset on hold. Ask for direction from the client's authorized contact for legal, privacy or communications matters. File the decision with the final approval materials.
Turn the checklist into a production gate
Run through the checklist at discovery and again before production. Come back to it before publication. Pick out the personal elements in the brief and the source files. Then set out the approved use and who has access to which tools. Settle retention and deletion. Name the stakeholders who can authorize release. In sensitive sectors, schedule escalation early enough that the production timeline stays safe.
By the end of the review you should have a written list. It records the decisions and the open questions, and the approved uses. Access controls, retention actions and release approvals go on it too. Include the project owner and the marketing or communications lead. The production team should be included as well, and so should the client contacts responsible for approval. Before publication, legal or privacy counsel should confirm any conclusions that depend on the jurisdiction.
We open our projects with discovery and a close look at the audience. Industry research is part of that start, and so is a communication strategy cut to fit the client. Bring along the AI video workflow you're proposing. Bring the source materials and the publication plan too, and the contacts who will approve the work.
Book an AI video compliance review. Send over your project scope and the channels you have in mind. Next comes a focused discussion of the questions to resolve before identifiable material enters the production pipeline. Those questions cover personal data and approval, and access and retention.



