One last question and then I’ll drop it, because I still think the ambiguity and contradictions here need to be addressed.
I’d like @VladTheImplier to let me know how these examples would be handled under the current rule.
First, I already understand that an AI-generated video of Scarlett Johansson would be considered a deepfake and therefore not allowed. But by your own explanation, if that same AI model is clearly portraying Black Widow, would it be allowed?
https://rule34.xxx/index.php?page=post&s=view&id=17773730
The AI is obviously trained on Scarlett Johansson’s likeness as Black Widow, but according to your explanation it should be acceptable because it’s portraying the character rather than Scarlett Johansson herself.
Second:
https://rule34.xxx/index.php?page=post&s=view&id=17050002
In my opinion this custom 3D model doesn’t really resemble Scarlett Johansson very closely (no offense to the artist), but the tags identify it as Scarlett Johansson rather than a fictional character. Even though the resemblance is fairly weak, would this be banned because the apparent intent is to depict Scarlett Johansson?
The reason I’m asking is that these examples highlight why I don’t think “intent” is a workable standard. Intent isn’t always obvious, and sometimes there are multiple reasonable interpretations of the same piece of content. Different moderators, or even the same moderator on different days, could reach different conclusions.
That’s my main concern. The current wording of the rule still leaves enough room for interpretation that they could be used to justify banning almost any CG depiction of a character based on a real person’s likeness, or allowing it. Users shouldn’t have to predict how a moderator will interpret “intent,” or read through a 50+ post discussion to understand all of the nuances of what is and isn’t allowed. The rule itself should be clear enough that someone reading it can reasonably determine whether their content complies before they post it.
Edit:
So that I don’t sound like a contrarian, I went ahead and wrote up what I think the rule should look like. Since there’s so much to cover, I feel this deserves its own section in the rules.
Summary
Deepfakes and Non-Consensual Depictions
Banned: Any content that sexually depicts a real, identifiable person without their consent. This includes:
- CG animation, 3D renders, or 2D art where the clear and primary intent is to depict a specific real person rather than an established fictional character
- Real footage or images of a person altered or manipulated to depict them engaging in sexual acts they did not consent to
- AI-generated content that places a real person’s likeness in a sexual scenario
- Character ‘parodies’ that are designed to avoid this rule and depict real celebrities (The Billie Eyelash situation)
Allowed:
- Sexual content featuring established fictional characters, even if those characters are portrayed by or physically resemble real people, provided the content is clearly about the character and not the actor (Black Widow, not Scarlett Johansson)
- Fan art or renders of fictional characters that incidentally resemble real people
- Babecock PMVs, where real, unaltered images of a person are displayed alongside sexual content but the person themselves is not depicted engaging in any sexual act
Grey areas will be moderated on a case by case basis, but the key question is whether a reasonable person looking at the content would conclude it is primarily about a real person or primarily about a fictional character. Context including title, tags, costume, scenario, and framing all factor in. Intent stated by the poster will be considered but is not sufficient on its own.
One question the rules still need to answer: does AI-generated content depicting a real person as an established fictional character fall under banned or allowed? There are three reasonable positions: (1) it follows the same character framing rules as CG and 3D and is allowed if clearly about the character; (2) it is always banned regardless of character framing given how indistinguishable AI can be from real footage; or (3) AI is held to a stricter standard than CG and 3D specifically because of that realism concern, even if realism isn’t a factor for other formats. Which of these the rules intend needs a defined answer.
