Skip to content

AI production records for freelance AI artists

When the client's lawyer has a question about how the work was made, there is no one to forward it to. Your record is your answer.

Published July 22, 2026.

Freelance contracts have grown AI clauses. Clients ask you to disclose which tools you used, to warrant your process, and sometimes to accept liability if that warranty fails. An agency signing the same language has a legal department and a production system behind it; a freelancer has whatever they wrote down at the time. That asymmetry is the whole argument for keeping records, and it is why freelancers who keep them sign confidently.

The structural problem is that a freelancer's work is scattered across personal accounts on several tools, and those accounts are rented, not owned. Histories live behind subscriptions, and a tool's history is available only as long as the account exists and the vendor chooses to keep it, neither of which is in your control. A record you capture and hold yourself does not depend on any of that, which is the practical difference between owning your evidence and borrowing it.

Records are also a commercial asset, not just a defensive one. The client who comes back a year later asking for the same look as the spring campaign is really asking a records question, and the artist who can reopen the prompts, settings, and references from that project wins the brief without a week of rediscovery. Your archive of how you achieved past work is part of what a repeat client is paying for.

Finally, records let you be precise about what you promise. A record can show what you prompted, what references you supplied, what you changed, and when, and that none of it has been altered since. It cannot show what a model was trained on, and it cannot show that one reference caused one output, so keep your contractual promises inside what your record can back. The provenance guide below maps that boundary in detail.

Where documentation breaks down

  • Generation history lives in tool accounts the artist pays for monthly, and it disappears with the subscription, the account, or the vendor's retention decisions.
  • Contracts ask for warranties about process months after delivery, when the artist has done a dozen projects since and the details have blurred together.
  • Clients ask for prompts and settings at handoff, and gathering them after the fact from several tools costs unpaid hours.
  • Disputes about whether the final was reworked from the approved draft come down to whose memory sounds more confident, unless a record settles it.
  • The artist's own style archive is scattered across tools and folders, so repeat briefs mean re-finding a look instead of reopening it.

A documentation routine that holds up

  1. Capture at generation time: prompt, model and version, settings, and references, into a system you own rather than the tool's own history.
  2. Note which account and plan each generation ran under, and the date, because commercial questions later often turn on exactly that.
  3. Keep drafts and the final in one chain per deliverable, so the path from approved draft to delivered file stays visible.
  4. At handoff, deliver a per-asset summary with the files, and keep the full record for the follow-up questions.
  5. When a project closes, archive the record under the client's name, because that archive is where the next brief from them starts.

Frequently asked questions

A contract asks me to warrant that my outputs do not infringe. Can records help?

They help with the part you can actually speak to. A record shows your process: what you prompted, which references you supplied, and what you changed by hand. No record can show what a model's training data contained, so push back on language that asks you to warrant that, and keep your promises inside your own process.

The client wants my prompts. Do I have to hand them over?

That is a contract question, and artists answer it both ways depending on the relationship and the fee. The point is that it should be your decision, and it only is one if you still have the prompts when the request comes.

My tools show my full history already. Why keep my own record?

Because that history is available only for as long as the account exists and the vendor keeps it, and neither is in your control. A record you export and hold survives a lapsed subscription, a changed retention policy, and a discontinued product, which are all normal events over the life of a client relationship.

What do I show a new client who asks how I work?

A redacted record from a past project. It demonstrates a working process in a way no statement about your process can, and it quietly answers the diligence questions before they are asked.

How much time does this actually take per project?

Capturing at generation time costs minutes, because everything you need is on screen at that moment. Reconstructing the same information after the fact costs hours across several tools, and the result is still incomplete.

Related guides

  • Prompt and model documentation: a practical checklist for studios The exact capture fields, sized for one person to keep up with alongside the actual work.
  • AI provenance: what you can prove and what you cannot The boundary your contract warranties should stay inside, explained before you sign.
  • How to document AI-generated work for client delivery What the client's side of the table is being asked for, which explains what they will ask you for.
  • EU AI Act Article 50 for creative teams The transparency obligations that reach freelance work through client contracts.
  • What is an AI production log? An AI production log is the per-asset record of prompts, models, settings, references, edits, rights decisions, and approvals behind AI-assisted work. Here is the minimum useful structure and the limits of what that record can prove.
  • C2PA vs. AI workflow documentation C2PA and AI workflow documentation solve different parts of the provenance problem. C2PA protects signed assertions attached to media; workflow records preserve prompts, sources, decisions, rights, versions, and approvals around the file.
  • AI shot log template for image and video production A practical AI shot log template for prompts, models, references, versions, human edits, disclosure, and approval. Download the CSV and adapt the field guide to your production workflow.
  • How to record client approval for AI-generated work A client approval record for AI-generated work should identify the exact version, disclosed AI use, review scope, requested changes, approver, and time. This guide turns sign-off into evidence instead of an ambiguous email.
  • Guide editorial and verification policy How Behind The Workflow researches, sources, dates, updates, and corrects its AI production guides. Product facts use primary sources, legal claims link to official text, and machine-readable markup mirrors visible content.
  • AI production records for AI producers AI producers own the delivery questions nobody else can answer: which model, which prompt, which version, approved by whom. This guide sets out the records that let a producer answer from evidence instead of memory.
  • AI production records for creative directors Creative directors approve what ships, so questions about AI in the work come to them, in client meetings and in review rooms. This guide covers the records that let them answer precisely instead of vaguely.
  • AI production records for post-production teams In post, AI arrives as a hundred small fixes inside shots rather than one big generation, which is exactly what makes it hard to declare on a delivery form later. This guide covers records that map to shots, versions, and conforms.
  • AI production records for agencies Agencies carry contractual responsibility for AI use across every team, freelancer, and production partner on an account. This guide covers records that hold at agency scale, where the person who generated and the person who answers are never the same.

Every requirement in this guide is answered from a record kept while the work happens. BTW keeps that record for you.

Start your own record