Risk Checklist for Using AI in Client Work: Disclosure, Releases, and Archiving

Reading Time: 6 minutes

Published: January 12, 2026

Client work comes with deadlines, approvals, and delivery pressure. AI tools can speed up production, but they also bring new risks around rights, realism, and recordkeeping. If you lose track of permission or forget to document what changed, that can create problems later. This guide gives you a checklist you can follow on real jobs, plus short explanations to help you adapt it based on the situation.

A risk checklist for using AI in client work is a structured set of checks that helps you manage three major questions: (1) do you have permission for faces, voices, and assets; (2) do you owe clear disclosure to a client, viewer, or platform; (3) can you back up what you did later with records and archives. This checklist does not replace legal review or contracts. It is built for film and video work where AI affects real people, brands, or real-world claims.

Legal and client note: This is practical guidance, not legal advice. Laws and platform rules vary by country, contract, and union. If a project involves public figures, political topics, medical claims, or broadcast delivery, include legal review and documentation support in your budget. When in doubt, contact an entertainment lawyer.

Why AI risks matter in client video work

Film and video are built on controlled illusion, but client work depends on trust. A viewer might believe a synthetic shot is real. A client might assume all rights were cleared. A platform might flag content that feels manipulated. When you keep clean paperwork and good records, you can answer those questions clearly and avoid problems.

• Disclosure explains how AI was used when confusion is likely
• Releases and consent show you have permission for identity and likeness
• Archiving and provenance give you proof of what tools were used and what was approved
• Workflow timing helps you apply checks at the right moments before delivery

Why AI risk feels sharper in professional projects

When you’re working for a client, there’s more pressure to get things right the first time. Once a project is released, it’s often reused or reposted far beyond what you planned. Small mistakes or unclear boundaries can cause real problems later.

Client work spreads beyond your control

That short hero video might end up in a paid ad, an internal training reel, or a trade show loop. You can’t always control where it ends up. Releases and disclosures need to fit where the video is actually going, not just where it started.

AI mixes capture and creation

In the past, it was easy to tell what was filmed and what was edited. AI tools can blend those things in a single shot. That’s why it’s important to document what was real, what was generated, and what was changed.

Risk rises when meaning changes

Tools like cleanup or transcription are usually low risk. But if the AI makes it look like someone said something they didn’t, or creates a shot that feels like a real event, that raises red flags. FilmDaft’s EU AI Act deepfake disclosure guide can help you spot these situations.

The three lanes to track: disclosure, consent, and provenance

These ideas often overlap, but they solve different problems. Keeping them separate makes it easier to explain what happened, prove you had permission, and protect yourself and the client.

Disclosure answers: “What are we looking at?”

Disclosure is what you tell the client, the viewer, or the platform. It might be on-screen text, a note in the deliverables, or caption info. If the content could be mistaken for real, disclosure helps clear that up.

Consent answers: “Are we allowed to use this person’s identity?”

Consent is the permission you get in writing. Someone agreeing to be filmed doesn’t mean you can use AI to change their face or voice. If you’re altering their identity, your release needs to say that clearly.

Provenance answers: “Can we prove what happened?”

Provenance means keeping track of what tools you used, what versions shipped, and what got approved. If a client wants to reuse something later, or if a platform flags your content, your records help explain what happened.

How to run the checklist during a real client workflow

This checklist works best when you apply it throughout the project. These four points are where most AI decisions are made, and where small problems can be caught early.

Moment 1: Brief and bid

Start by setting expectations clearly. Make sure you know what AI is being used and how it could affect rights or realism.

• Step 1: Describe the planned AI use in plain terms
• Step 2: Note identity exposure (faces, voices, trademarks, private locations)
• Step 3: Decide who approves AI use and where that approval is saved

Moment 2: Production and asset intake

This is when you start gathering materials. It’s also when most legal mistakes happen if releases are skipped or misplaced.

• Step 1: Confirm what can leave the set or be uploaded
• Step 2: Confirm release coverage for everyone who appears or speaks
• Step 3: Start a simple asset log so you can track inputs and outputs

Moment 3: Post-production and review

As the edit comes together, it’s easy to forget which shots were altered or enhanced. Marking AI-generated or AI-altered material now saves time later.

• Step 1: Track changes to meaning, identity, or realism
• Step 2: Build a disclosure plan for each version (video, cutdowns, captions)
• Step 3: Save proof along the way, not just at the end

Moment 4: Delivery and archive

Once the work is done, your job is to package it clearly and archive what matters. That way, you can prove what happened even if someone else handles it later.

• Step 1: Write clear delivery notes (avoid vague terms like “AI used”)
• Step 2: Export a proof pack with contracts, releases, AI logs, and versions
• Step 3: Store it securely with limited access

The practical risk checklist (run before delivery)

Use this list to double-check everything before the final export. It’s built to catch the most common points where things go wrong.

1. Describe the AI use clearly and briefly
2. Check for identity exposure (face, voice, name, performance)
3. Get consent if the AI created a digital replica
4. Confirm all talent or model releases are signed
5. Make sure you have location or property releases if needed
6. Identify third-party assets and confirm usage rights
7. Check if client material was uploaded or shared externally
8. Decide who needs disclosure (client, public, platform)
9. Place disclosure in a visible, persistent spot
10. Test the riskiest clip on someone outside the team
11. Record which versions shipped and where
12. Archive a proof pack that matches the risk level

Disclosure in practice

Disclosure helps prevent misunderstandings. When something looks real but isn’t, a simple label keeps your work transparent and trustworthy.

Client-facing disclosure

Even if the client doesn’t ask about AI use, adding a short note about how it was used can help future vendors or reuse scenarios.

Viewer-facing disclosure

If your footage uses visual styles that look like real news or documentary content, be more careful with labeling. What feels obvious to you might confuse others.

A clear method to follow

• Mark every shot that involved AI generation or transformation
• Ask: could this be mistaken for real footage?
• Decide where the disclosure should go (credits, captions, on-screen text)
• Upload and double-check that the label stays visible

Releases and consent

Releases only matter when something goes wrong — but by then, it’s too late to fix them. Use this moment to build solid habits and avoid guesswork.

Match the release to the role

Use a talent release for performers and a model release for regular people who appear on camera. Each covers different platforms and use cases.

Digital replicas need clear approval

If the AI is generating speech, changing faces, or extending performances, your release must mention that in plain terms. Be specific about use cases, platforms, duration, and reuse.

Standardize your release pack

• Model or talent release for anyone on camera
• Written permission for voice-only roles
• Separate consent for digital replicas
• Location or property release for recognizable places
• Client approvals for final delivery and AI use

Archiving and provenance

Months from now, someone may ask what tools you used, who gave approval, or whether something can be reused. A good archive helps you answer quickly and safely.

Keep a clean proof pack

Think of your archive as a folder that tells the full story. It doesn’t need every raw file — just the ones that prove the key decisions and approvals.

Protect sensitive info

Releases often contain names, contact details, and signatures. Store them in a system with access limits and clear responsibility.

What to include in your archive

• Contracts, scope documents, and change orders
• Signed releases and digital replica consents
• Tool names and processing locations
• AI prompts and settings (if relevant to results)
• Client approval trail for AI use and final cut
• Exported files with delivery notes

Common misunderstandings

Most problems come from reasonable assumptions that don’t hold up under review. These are some of the most common.

Filming someone doesn’t mean you can generate new material

Consent to appear in a video is not consent for AI to generate new speech or facial expressions. Treat that as a separate step.

Metadata isn’t enough

Captions or on-screen text work better than metadata, which is often stripped or hidden after upload.

Client ownership doesn’t remove your risk

If your contract says you cleared rights, you’re still responsible even if the client owns the file.

Generic AI shots can still imply facts

Footage of “people working” can look like a real company or real event. If it feels like documentary, treat it as high risk.

When to get help

Some jobs are too sensitive to figure out on the fly. Pausing to get expert input can save your schedule and your budget.

High-risk projects

If you’re working with public figures, health topics, politics, or minors, pause and verify disclosure, consent, and delivery records before anything is released.

Unclear asset ownership

If you or your client can’t prove who owns something, don’t build AI on top of it.

Use legal review as a tool

FilmDaft’s entertainment lawyer guide explains when legal input is worth the cost.

Summing Up

A clear AI risk checklist helps you manage disclosure, consent, and proof. Use it throughout the job, not just at the end. Keep your release pack simple but specific. Put disclosure in places where it survives reposts. Store a proof pack that answers the hard questions. When the project is sensitive, treat legal review like any other production tool — not something extra.