Generative AI - Policy?

posted by: spandexpartyadmin

I think AI-generated content on Spandex Party has now grown to a point where we need to take some action. So this post got a bit lengthy, please bear with me. Pretty much every second or third report coming through the reporting tool at the moment is related in some way to AI misuse. We don't even have an "AI content" reporting category yet, so people are reporting it under things like "off-spandex" or "fake". GENERATIVE AI VS AI-ASSISTED EDITING I think it's important to distinguish between generative AI and the kind of AI-assisted editing that has become part of normal photo editing. By generative AI, I mean images where AI creates or substantially replaces the person, body, clothing, setting, etc. That's quite different from using AI-powered tools to improve lighting, reduce noise, sharpen an image, remove something from the background, or make other relatively small corrections to a real photo. WE'VE HAD A SIMILAR DISCUSSION BEFORE Some time ago we had a similar discussion, but it was about nudity and pictures that don't show any spandex. The consensus was basically: if someone has a profile with lots of spandex pictures and throws in the occasional non-spandex picture, that's okay. But if a profile consists almost entirely of non-spandex pictures and close-ups of someone's cock or ass, then that's not really what Spandex Party is for. WHY I THINK AI IS DIFFERENT With generative AI, though, I don't think we can apply quite the same approach. A generated image ultimately isn't a picture of the profile owner. You could argue that someone is uploading a representation of their ultimate fantasy, or something they would like to experience. But that very quickly becomes difficult for us to define, check and moderate consistently. For that reason, I would personally prefer us NOT to allow fully or substantially AI-generated profile pictures at all. I share the same view as others here: profile pictures should fundamentally be about the profile owner. WHAT COULD WE DO? A few practical things come to mind: - Add a clear "DO NOT UPLOAD AI-GENERATED CONTENT" explanation to the upload modal. - Potentially add a checkbox or declaration around AI use when uploading a picture. - Add "AI-generated content" as a reason in the reporting feature. THE GREY AREA: AI ENHANCEMENTS I think this is the difficult part. I'm sure quite a few people already use tools that include AI to slightly improve their photos. Personally, I think minor fixes and enhancements should remain okay. I don't think we want to start investigating whether somebody used an AI denoiser, removed something from the background, improved the lighting, sharpened a picture, or made another small correction. Technically, external tools may sometimes be able to detect AI manipulation, but the last thing I want is for this to turn into an "AI witch hunt" with endless reports and arguments over whether a perfectly normal-looking photo was edited by AI. WHERE I WOULD DRAW THE LINE For me, the useful distinction would be: - AI-assisted editing of a genuine photo of yourself: okay (but maybe with a AI badge for labelling) - AI-generated or substantially AI-created images presented as profile pictures: not okay. What do you say? what would you change or add? To me this feels like a rule we could explain relatively clearly and, more importantly, moderate reasonably consistently.

posted by: spandexpartyadmin

I think AI-generated content on Spandex Party has now grown to a point where we need to take some action. So this post got a bit lengthy, please bear with me. Pretty much every second or third report coming through the reporting tool at the moment is related in some way to AI misuse. We don't even have an "AI content" reporting category yet, so people are reporting it under things like "off-spandex" or "fake". GENERATIVE AI VS AI-ASSISTED EDITING I think it's important to distinguish between generative AI and the kind of AI-assisted editing that has become part of normal photo editing. By generative AI, I mean images where AI creates or substantially replaces the person, body, clothing, setting, etc. That's quite different from using AI-powered tools to improve lighting, reduce noise, sharpen an image, remove something from the background, or make other relatively small corrections to a real photo. WE'VE HAD A SIMILAR DISCUSSION BEFORE Some time ago we had a similar discussion, but it was about nudity and pictures that don't show any spandex. The consensus was basically: if someone has a profile with lots of spandex pictures and throws in the occasional non-spandex picture, that's okay. But if a profile consists almost entirely of non-spandex pictures and close-ups of someone's cock or ass, then that's not really what Spandex Party is for. WHY I THINK AI IS DIFFERENT With generative AI, though, I don't think we can apply quite the same approach. A generated image ultimately isn't a picture of the profile owner. You could argue that someone is uploading a representation of their ultimate fantasy, or something they would like to experience. But that very quickly becomes difficult for us to define, check and moderate consistently. For that reason, I would personally prefer us NOT to allow fully or substantially AI-generated profile pictures at all. I share the same view as others here: profile pictures should fundamentally be about the profile owner. WHAT COULD WE DO? A few practical things come to mind: - Add a clear "DO NOT UPLOAD AI-GENERATED CONTENT" explanation to the upload modal. - Potentially add a checkbox or declaration around AI use when uploading a picture. - Add "AI-generated content" as a reason in the reporting feature. THE GREY AREA: AI ENHANCEMENTS I think this is the difficult part. I'm sure quite a few people already use tools that include AI to slightly improve their photos. Personally, I think minor fixes and enhancements should remain okay. I don't think we want to start investigating whether somebody used an AI denoiser, removed something from the background, improved the lighting, sharpened a picture, or made another small correction. Technically, external tools may sometimes be able to detect AI manipulation, but the last thing I want is for this to turn into an "AI witch hunt" with endless reports and arguments over whether a perfectly normal-looking photo was edited by AI. WHERE I WOULD DRAW THE LINE For me, the useful distinction would be: - AI-assisted editing of a genuine photo of yourself: okay (but maybe with a AI badge for labelling) - AI-generated or substantially AI-created images presented as profile pictures: not okay. What do you say? what would you change or add? To me this feels like a rule we could explain relatively clearly and, more importantly, moderate reasonably consistently.