VRChat Needs an Enforced AI Tag

(This took an extended amount of time to write then originally intended. The original plan was to provide a larger comment utilizing the “other” selection in the VRChat World Authorship Marking poll. Though as I continued to write and share a combination of my perspective and frustration towards the proposed “solution”. I realized my message became much more then a simple ‘Yes’ or ‘No’ answer, but evolved more into addressing some of the core problems VRChat in of itself failed to properly address the topic of the big elephant in the room; AI worlds. As well failing to properly address the outcry from world creators and the many players I have come across through the platform who shared, regardless of where they stood on the issue, a pretty lateral response to what they considered was a bare minimum bandaid solution: An Enforced AI Tag.

With that being said, allow me to share what I decided to spend a few hours of my time writing on a Wednesday night then spend time within a virtual community of people that I love.)

No. VRChat needs an enforced AI marker so users have better expectations of the worlds they explore, how they engage with creators and how they search for worlds. This is especially important when creators offer their services and users have no way of knowing what tools they use in their work. Sometimes a world utilizes AI in a neat or interesting way. Sometimes it doesn’t. I am not strictly against the use of AI, but I am firmly in favor of enforced tags and labeling. Users should be able to make informed decisions about the content they engage with and understand what creators are capable of achieving with AI in their worlds.

If the concern is targeted harassment, criticism, hate, and cyberbullying, those are social problems that exist regardless of whether a tag is present. People are already capable of identifying AI-generated content, and they can easily find creators across the web. Refusing to implement a proper, enforced labeling system creates far more problems for everyone, regardless of whether they use AI in their world creation.

An optional, unenforced Human-authored mark is not a solution. Creators can already add text to their world thumbnails or descriptions to identify their work as human-authored. Anyone can do this without any meaningful support from VRChat. This isn’t even a bandaid solution, and I think it’s insulting to suggest it to world creators who take pride in building their worlds without AI tools. It puts the responsibility entirely on creators who want to distinguish their work, while doing nothing to establish a consistent and reliable labeling system across the platform. It creates unnecessary friction and only adds to the frustration of a community asking for a straightforward solution.

Imagine if this was the final straw. Imagine if half, if not more, of the most popular worlds were delisted or removed tomorrow because their creators no longer wanted to support your platform. Imagine a platform where the only worlds left were those created with AI. How many players would stay? How many would leave? These are not outcomes VRChat should be willing to risk over a labeling system that could have been implemented from the beginning.

Again, I am not against AI usage. However, I have a serious problem with poor decision-making, a lack of foresight, and the creation of a community problem that could persist across VRChat indefinitely if it is not properly addressed and taken seriously. From my perspective, VRChat does not have a strong understanding of the community’s perspective on this issue. At best, there appears to be only a surface-level understanding of the playerbase, with limited direct communication and professional relationships with the people who actually create and maintain the content that makes this platform worth using.

Based on what Tupper has shared and the decisions VRChat has felt comfortable moving forward with, I get the impression that VRChat is unwilling to provide a proper solution or even a reasonable bandaid approach. The proposed Like system is one example. It works against creators, can easily be abused regardless of the safeguards put in place, and has the potential to cause immediate problems upon implementation and much more serious issues in the long term. These kinds of decisions make me question whether VRChat fully understands the consequences of its actions or the people who will ultimately bear the cost of them.

VRChat needs a direct line of communication with creators and community leaders to establish a stronger working relationship with the playerbase. The people creating the worlds, avatars, and experiences that keep users engaged should not feel ignored when raising legitimate concerns about the future of the platform. There needs to be a meaningful process for listening to these concerns before decisions are made that could permanently damage the relationship between VRChat and its creators.*

*Small clarification on my part, I know Tupper said they engaged with creators in one-on-one conversations on the topic. My point was that there is no meaningful process for listening to concerns across a open conversation where creators, community leaders and the extended community of passionate VRChat players who are willing to contribute to that conversation.

All I see right now is an approaching community disaster that could become far more damaging to VRChat if it is not addressed quickly and properly. There needs to be a firm commitment to an enforced AI labeling system that acknowledges the concerns of the large portion of the community that does not want AI-generated or AI-assisted content presented without clear disclosure. Users deserve transparency, and creators deserve a consistent system that allows their work to be understood on its own terms. This should not be about forcing everyone to agree with or oppose AI. It should be about giving people the information they need to make their own decisions.

From my perspective, the current direction feels overwhelmingly in favor of users who want to utilize AI as a tool, while completely disregarding those who do not want it to become an undisclosed part of their experiences. As someone who does not want to see creators, their worlds, and their communities leave VRChat because the platform refuses to provide a proper solution, I find this deeply infuriating. The failure to understand why an enforced AI tag was the simplest and most reasonable bandaid solution is concerning in itself. It gives the impression that VRChat is unwilling to properly address this issue, let alone acknowledge the scale of the problem it could become.

VRChat needs to take this seriously before frustration turns into action and creators begin leaving the platform. The longer this is ignored, the more difficult it will be to repair the damage. An enforced AI marker was never intended to be a permanent solution. It was a proposed bandaid that would provide greater transparency in the short term while giving VRChat more time to work toward a permanent solution.

An enforced AI marker is a basic transparency measure that benefits users, respects creators, and gives the community the ability to make informed decisions about the content it engages with. Refusing to implement one risks creating a division that VRChat may not be able to undo. I genuinely believe this is a preventable problem, but only if VRChat is willing to listen, acknowledge the concerns being raised, and make a decision that reflects the interests of its entire community rather than just one side of it.

16 Likes

I also come from the feeling of. A.I itself isn’t inherently bad or aweful. But just like profanity and horror are filters, so should A.I made stuff. I’d like the option to just, not see it at all. But I feel like alot of the devs are very positive to A.I and actually like it. So they don’t want to shun it as they themselves like it, Which to be clear. Giving it a tag is not shunning A.I. Again, just like the NSFW filter, or Profanity filter, or the horror/gore filter. It’s not shunning it, It’s informing users that, that content exists in this world/avatar and the user should be allowed the right and option to not view it.

And if it becomes a case of, people start not tagging their work of using A.I properly, well there’s a report system. The dev’s can take a look at the world, and see if the report is warranted. Just like with NSFW content.

There is no reason for there NOT to be an A.I tag, other than. The devs like A.I and feel like giving it a tag, is shunning it.

4 Likes

That was a very elaborate take. Rather than responding only from my own personal stance, I’ll jump straight to the conclusion I arrived at after more than an hour of discussion with friends and creators.

The core conflict here is between pragmatism and creativity.

Personally, I’m a pragmatist on this issue: if it looks good and works well, I don’t care how it was made. The rating system will sort everything out.

VRChat has adopted our perspective they provide a “created by a human” label, implicitly assuming that everyone else might be using AI. Their rating system doesn’t necessarily reward transparency or manual labor.

For artists and world creators, shifting the burden of proof onto them essentially saying, “prove that you’re human” feels insulting, as it devalues ​​the fundamental standard of human creativity.

The same logic applies to users: some care primarily about the final result, while others value the process and transparency.

VRChat has gone through a similar community schism before, back when they introduced EAC. History has a way of repeating itself…

The platform could have implemented a detailed specification for tools labeling things like “AI-generated assets,” “AI-written code,” or “AI-generated voice acting” instead of forcing a demeaning, binary choice between the stigma of “I used AI” and the defensive claim “I am a human.” But the question remains: is it even possible to build a reliable moderation system for such tags? It’s doubtful.


I’m a world creator myself; AI doesn’t offend me, even though I personally prefer manual labor. That’s just the world we live in we have to adapt.

8 Likes

Honestly having multiple tags for different levels of AI assistance would be beneficial. The debate itself is never going away since Pandora’s box was opened a few years ago. I absolutely despise AI, but I’m only one person and know that everyone has different levels of what they’re okay with.

I also completely agree with OP that failing to implement a bandaid solution for now is not a good look and honestly should’ve been implemented when 95% of everyone that responded said they wanted it.

3 Likes

Frankly, I have read most of your comments on this topic throughout several different threads, and it seems like you are out of touch. I don’t believe you understand the majority of the player-base, nor do I believe you are an expert on anything except narcissism.
Ai as a whole may not be bad, but any “generative” ai can be assumed to be made using stolen assets/art from other creators. As that’s exactly what it is.

Generative ai as a whole is just horrible. Bad for the environment, bad for business, and bad for other creators and artists. To prove this point, an ai generated movie was recently released in china, and lost money, because noone wants to watch or interact with it. A human made movie that had a budget of $200 made more than it because they did not use ai.
If the movie did not disclose its usage of ai, it probably would have made more than it did, however I believe that would have been morally wrong and legally questionable.
People care about this topic. If i find something I like online, whether its art or animation or a game or something else I will like it. If i then later find out that thing used AI in any stage of its development, I will remove my like retroactively. The reason is that I have seen the environmental impact that AI has on the world, and it is NOT GOOD. AI has directly resulted in our lives getting harder, people dying, and people getting dumber the more they use it(no offense).
It is my moral and religious obligation to denounce AI.
Anyone who is defending ai because of “pragmatism” is either lying to the world, or lying to themselves.

11 Likes

Ok so I hate to say it but every time somone says this they dont address the core problem that the VRC team pointed out in their original post.

HOW?

How do you enforce a tag that you cant easily determine? You can’t check compiled code and even if you could you can’t accurately determine the level of AI use. A well trained eye can spot AI art but thats not always the case and that takes both time and resources that VRC does not have.

The tag is nothing more than an honor system — an honor system that the bad actors won’t willingly use anyway. This was all addressed in the original VRC post mentioned here:

The reality you guys have to accept is that there is no scalable solution to this problem. Hell, even Steam is struggling to enforce these sorts of policies and I don’t think I need to explain that Steams moderation dwarfs that of VRChat.

If you want an AI tag, then you need to help the staff of VRC come to a conclusion on how they can enforce it effectively without overdrawing from their limited resources. The likes and dislikes are currently the best option as they leave the determination in the hands of the community. The community is a free source of quality control.

I personally was thinking about a potential “User Applied Tags” that would allow the community to determine a few tags that match the world, an AI one (or several) being included to allow the community to determine if something is elegable, but even that can cause witch hunts.

Long story short there is no good solution and the fact that the community is so willing to review bomb the project rather than try and address the concerns shows that there isn’t enough knowledge of the requirements of this system. They explained the situation the best they could in a series of posts, and I have yet to see someone actually address them instead of just calling them lazy or anti-artist.

4 Likes

Devs already said no AI filter, so giving them advice at this stage seems rather pointless. If, and when, they change their tune, we can discuss that.

Meanwhile, I will keep speaking up, which many people are, (not “review bombing”) for my belief in human-made content first, and fighting back against this artificial corruption and unathorized bastardization of the very soul of humanity, art.

I won’t allow those furries and others who secretly/not-so-secretly enjoy/generate questionable content to have their way. I WILL be loud and speak up for what I believe in, it’s one of the few good traits I have, or so I’ve been told.

On a side note, happy Blue Texas day! :slight_smile: Hope the democrats get off their asses and come after AI companies.

1 Like

Have you, by chance, looked at the Steam review charts? Yes, many people who hold your views are, in fact, review bombing the project.

As for not taking advice: just because they decided against an unsustainable practice doesn’t mean they are against ALL advice or that they reject potential systems that could meet your desires while also being sustainable. That’s why I decided to shoot my idea for community driven tags.

Also, you haven’t actually acknowledged any of the points I made. Great job turning this into a political issue, though, especially when a large portion of conservatives, myself included, are strongly against AI generation.

(As a quick side note: conservatives aren’t always Republicans.)

3 Likes

I am not obligated to! :) As somebody who is queer and anti-ai however, it is, as ever, my obligation to speak up and try to create change, even if I can’t change things on my own, I’d rather try. All the more reason to, especially when those who were financially backing these AI companies were largely right wing. Those who would erase my, also queer, friends.

Memories of those precious friends, mean the world to me. And they keep me going. My motivation for breathing another day…

So, I won’t quiet down.

And I especially won’t stand idly by when you or anyone else tries to pollute the Earth further with your slop-making data centers!

Say no to AI. Say yes to our beloved queer friends and creators in VRChat.

:blue_heart:

Oh good heavens… Those financially backing AI companies do not care for you or me; they only care about putting money in their pocket. You’re acting like this is a political issue. It’s especially not political when it comes to VRChat’s AI issue.

I also never told you to quiet down. I’m on the verge of doing so because you won’t acknowledge the original point. My intention was for you to be just as vociferous but to make sure you do so logically. Don’t just scream to scream. Help VRC solve the problem.

We all agree AI is a problem. We may not agree to what degree, but regardless, we all want change. With that said, give options. They already stated why they cannot do a tag, so take that and try to come up with a system that doesn’t hit those roadblocks. Hey, maybe you come up with a way that the tag can work! My point is that you won’t if you do not address the concerns of the original post.

1 Like

This is so pretentious I have no idea how to address this.

No one is targeting you, nor is anyone harming you specifically in this conversation about AI. You are not being targeted for being queer, no one even brought that up.

People hate AI, I hate generative AI. However, literally this individual just asked how do you do it?
And you just exploded on them with vitriol.
Moderation needs employees to verify reports, or a company outsourced to moderate those reports. How is a company to pay that person to moderate? How are those rules set? How do you verify something was created with AI?

Are you specifically @Mobilized_Suit going to fund the salaries, and insurance costs to fund these employees? If not, how are the people that will be sitting 8+ hours a day through reports supposed to handle feeding themselves or their families? Will it be volunteer?
If it’s volunteer, who manages the volunteers? How do they manage them? What happens if one of those volunteers is a malicious actor or causes problem within the company?

I get that people are angry, if I am being honest, so am I.
However, you don’t realize the amount of things that need to be balanced in order to make ANYTHING work. Period.

It’s not impossible, it just takes time and real solutions designed by real people over a period of time.

1 Like

A real answer to this question is that I think it needs to follow the same ideas and ideals as AO3.

AO3 is one of the largest fanfiction/writing sites in existence, its user-base is much larger than VRChat. They have tags, and warnings, for those who are or are not interested in certain things as they allow practically anything to be written.

You don’t like something? Too bad, add it to your block list. You will then never see it again, but it still exists on the platform.
Sometimes the bad actors or people too lazy to tag falls through, but, I feel like that is what Community Labs would be best for.

Implementing it is a bandaid to the gushing wound, but, one that can be healed over time.

Coming from a more coding-focused side, I’ve used so much AI in the last couple of years for shader development and udon scripts. Just asking the AI whenever I encounter any roadblocks (which was often, lol). But I’ve always hit limits where the AI can’t answer and just makes stuff up or it randomly deletes a part of the logic. The amount of time I have wasted with “please fix it fr this time” is so high I probably don’t want to know. And once you give in to the laziness and just copy an entire AI-generated script, you’re f-ed. You have no clue how “your” code works, and once a new issue arises, you don’t know how/where to fix it yourself and have to ask the AI again. The codebase doesn’t belong to you anymore at this point. It took away much of the joy for projects of mine where I relied on it. I’m at the point where I only use AI to find certain functions instead of crawling through documentation or copying small code snippets it slops out.
If you ever want to get into scripting or shaders, read a book or tutorials. I’ve gone the path of generating shaders and “learning” from them, but I would have been sooo much faster by just learning from a well-made source.
I’m reading the (now free) gamemath book atm and can heavily recommend it.

Exactly. Thats why I recomended community driven tags. It could be a useful way for the community to drive moderation rather than requiring VRC to try and keep up. There would have to be a way to combat it if users decided to try and bomb a world but if we can figure that out I’m sure it would be more practical than what people are currently asking for.

I am not going to argue in good faith with a self-identified conservative. Nor is it my job to figure out how a company which is profiting directly off of its users through fomo accessories and cosmetics along with monthly subscriptions, how to fund the fix many people are requesting. Good day. <3

“I want to complain without giving constructive criticism, and for some reason, my queerness makes me unable to have like opinions as someone else.”

Please note I did not once belittle your political stance, but rather attempted to bring the conversation away from politics. So what if I’m conservative? Do we not both hold the same stance here? If we do, why can’t we attempt to come to an agreeable system? You’ve given no input, but you are entirely willing to belittle those who are attempting to solve the problem.

If it is not your job to help solve the problem, you have no right to complain about the problem. VRChat is a community-driven platform. If you do not wish to take part in helping this community, you have no right to be a part of this community.

I think this would be an unintentional net positive for VRChat. Unintentional in that I’m going to assume that the point of leaving a poor review on steam is not to specifically encourage random people to avoid playing VRChat. VRChats new user experience currently sucks, and I dunno about most people but when my friend suggested I take a look at VRChat I didn’t look at the steam reviews. So in that respect I kinda avoided the new user experience of not knowing anything and likely joining a public black cat.

For this AI flagging discussion I do wonder if any externally operated websites are making a list. Personally I’d be interested if anyone’s looking through the new worlds list and picking out anything that isn’t either a lazy asset assembly or AI generated lazy asset assembly.

That’s fair. I guess the effect isnt really on the user base or even new users for that matter.

Either way, it’s frusterating seeing that many people throw a hissy fit when this could have all been avoided simply by having these individuals read the post they are protesting.

As for a list, I think that would be unsustainable without ripping the worlds and using bots. It just isn’t really something anoyone logically can keep up with.

I would like to add: Part of the core reason that people wish to not engage with worlds that use varying levels of AI is due to how AI, in a lot of cases, removes the human element from it. If we are inspired, influenced, and otherwise live inside of these worlds, it stands to reason that they shape our experiences.

People wish to not have to be influenced by content that was made with AI due to how that fundamentally influences our culture.

I agree that the sentiment against AI users is bad, and maybe that will change, just like the sentiment towards horror creators isn’t bad. That will be difficult, as AI users often wish that their work was respected on the same level as human-created content and they continue to shove it in people’s faces (with one user claiming that their extensive use of AI counts as “human-authored”), but fear of harassment is not a reason to not implement these features. It’s a basic right to protect our minds from negative influence.

Statistics in reporting. If a world suddenly gets a surge of reports that is disconnected from its heat level, especially if you check whether or not that reporting user even entered the world, there is a very good chance that is a brigade.

If reporting seems more natural, not bunched up, and correlated with world popularity + verifying those users actually engaged with the world in some way, those reports can be verified just with basic math.

Not to mention, there is an incentive for AI users to tag their stuff properly. If people who don’t like AI can simply block and ignore that content from their local client entirely, it means they literally don’t engage with it at all. That’s literally what most people want, and VRChat refuses to do that.

VRChat absolutely has the capability to make systems like this. It is likely they already have systems like this to detect brigading or other negative reporting patterns.

I totally get that people don’t understand how, but there are probably more ways than this to enforce it. I think it’s an insult to humans to suggest that we have no capability to solve problems like this.

The massive, massive hate towards AI users is them posturing that their works, especially slop works, are equivalent to human authored work. It’s their insistence on shoving AI into every aspect of society regardless if people consent or not.

Even me, as someone who uses Claude every day and pays Anthropic (a garbage company) $20/mo because I see value in the tool, sees the value in these labels and I think everyone should have a right to avoid AI content.