A common way to describe an AI tool on a show.
Da-M-I-C-Us is a classification scheme from the VES Technology Committee’s generative AI work: five dimensions — Data, Model, Infrastructure, Code and Use — that let a studio, a vendor and a client say precisely what an AI deployment is before anyone argues about whether it is allowed. The factual classification stays separate from the policy that judges it.
Da-M-I-C-Us v1.0, draft 2 · working material, not a ratified VES standard.
WD · Web-collected data Rights not individually cleared 
OM · Open model Permissively licensed weights 
SI · Self-managed Inference on your own machines 
OC · Open code Surrounding software is open 
Use · unset A capability class — no verdict attaches The five dimensions
| Dimension | Codes | The question it answers | |
|---|---|---|---|
 | Data (Da) | PD LD ND WD UD | What was the model trained on, and are those rights cleared for this use? |
| Model (M) | PM OM CM RM | Who owns or licenses the weights, and on what terms are they available? |
| Infrastructure (I) | SI EI DI MI | Where does inference run — your machines, the edge, or someone else’s service? |
| Code (C) | PC OC CC RC UC | What is the licence status of the surrounding software and pipeline? |
 | Use (Us) | BU PU IU DU | What is it actually being used for on this show? Leave it unset and you have a capability class, not a deployment. |
A verdict attaches to the full five-code string, never to a single code. Four-part strings (Use left unset) describe a capability; the fifth part makes it a deployment that policy can judge.
Prototype tools
Classifier
Answer the questions and the five-code string builds as you go, with a printable classification card at the end.
Normative · r10Policy Maker
Build an organisation’s rule list — permitted, by approval, refused — and export it as JSON the other tools can read.
Exploration shellCompare
What it is on the left, what the organisation permits on the right, and the verdict in the circle between them.
Prototypes from the working group, offline single files. Verdicts shown use draft 2’s illustrative internal policy; an organisation substitutes its own.
Where the thinking came from
The classification work follows a run of public sessions on what generative AI is actually doing inside visual effects — in artists’ hands, in the software, and in the contracts around both.
- 01 Real Applications of Generative Artificial Intelligence in Visual Effects What is in production today, shown by the people running it.
- 02 Generative Artificial Intelligence in VFX Software: Now and In the Future How the tool vendors are building it into the pipeline.
- 03 How Generative Artificial Intelligence Might Affect Visual Effects Now and In the Future The longer view: craft, employment and the shape of the work.
Two documents, one scheme
Start with the one-page sheet if you need to classify something this afternoon. Go to the full document for the definitions, the matching rules and the worked examples — and to leave comments for the working group.
PD Owned data 
PM Owned weights SI Self-managed 
PC Proprietary code An in-house model on owned data — the same notation, a very different conversation.