Power consumption is more of a concern when using company servers that you don't control, I agree that using a model locally on your own machine would be (probably) harmless. And by the same logic, training it on only data that you own should also be harmless.
In regard to tagging though, I would suspect that Itch wants to keep things as simple as possible. If you really need a disclaimer then you can add it to your page, each developer can be as specific as they want about their process.
Viewing post in Generative AI Disclosure tagging
Even when you use company server it is still more efficient as it does not really matter where you use energy - on your pc or on commercial server. (In fact, I believe commercial servers are more energy efficient). Because those networks use a lot of energy in a short period of time but alternative is to use average amount of energy for much longer period of time. Of course it is not always like that and just to be clear: I do not have exact numbers but it is how I assume the situation
But when we are talking about energy I always think people who really care about energy that much should not use x86 processors and services like Steam or Itch as they run their software on big commercial servers. Because where to put the line?
And by the same logic, training it on only data that you own should also be harmless.
And yet, I usually see all generative ai treated the same way in these discussions and with the handling of the tech. There is no differenciation at this point about any legally non-ambigous or ethically "clean" solutions. It is ai or not ai. And that is why I am lamenting that arguments against the tech should be solid and future proof instead of shortsighted and emotional.
With the current/initial filters provided there is also only ai yes or no with no practical information if it is content or just the code (yet?). Unlike artists, developers embraced the tech. Or at least that is my impression. It is the developers that create code with the tech. But with art, it is usually not the artists that use it to create more art, but people that could not draw at all to create images. Maybe it is, because in software it is a common concept to reuse other's code from templates, examples, and of course your own previous works and adapt it to new situations. In the visual arts that is frowned upon as "tracing".
There are many artists using AI. Also many painters, that generate references with it, or animate their own paintings in digital, and many mixed media artists, or hybrid artists. They all can draw and paint, as myself.
In one year or less, as public domain fully public sourced models appear and people identify the ones that are ethical, everyone will use it, except a few.
I marked my game with the AI tag, gladly. I hope the name changes to "ai usage", but it will be obvious that my work is a mix of traditional and generated skills as soon as people see it, so I don't really care. It's impossible to make a half decent game solely on AI.
The confusion about licenses, copyright, and polemizing on all this subject, fueled by the media, has damaged everyone. I understand the hate about the flooding of crap, but I hope it also becomes an old issue (exactly as when people got tired of DSL cameras and stopped annoying everyone making photos all day long).