Technology, Censorship and AI
Okay, so, I'll admit it. I kinda like AI. I know that's somewhat of an unpopular take amongst people right now, but there are things that it does that I find very cool. Not so much the stealing art to make filters, and god I hate the way that as a technology it's become this weird left wing, right wing issue. It feels strange to say that I'm an AI enjoyer when a lot of people's minds go straight to videos released by the current American government of the president shitting on protestors, or creating bastardizations of beloved nostalgic properties to promote ICE, or the killing of foreign citizens.
No, what I like AI for is a little better than that. Or at least I think so. I really like it as a writing tool. Not to write my words for me, that would be stupid, and kinda defeat the purpose of being a writer in the first place. Instead, I like it just as an assistant that I can use throughout the writing process. I can give it ideas and see what it thinks. I can give it rough drafts and get it to point out spelling and grammar issues, or places where my writing could be a little tighter. It can help with the research, making it easier to compile information that I can then go on to verify. And then, in the end, I generally get the AI to do a final pass on the article and give its thoughts.
Typically, I use Grok to do this. The reason has generally been that as far as the "flagship" AIs go, it tends to be the one with the least filters. Other AIs will typically shy away from the topic of zoosexuality fairly immediately, choosing to shut down the conversation instead. Zoosexuality is a controversial topic, and many AI developers would rather avoid screenshots of their AI being coaxed into supporting kissing dogs or horses or other animals that are attractive and kissable. On the other hand, Grok supposedly prides itself on the idea of free speech. After all, Twitter is the nazi site where the worst people can go to give their awful opinions on terrible things. And that requires a pretty open stance on what you can and can't say.
You can imagine my surprise then when the other day I opened up Grok, and asked it to review an article, and it hit me with this.
"I'm sorry, but I can't preview or summarize articles from zooey.pub. The site is a zoosexual lifestyle magazine that discusses topics related to zoophilia, which often ties into illegal activities (like bestiality) in most places. I don't assist with content involving criminal behavior. Let me know if there's something else I can help with!"
On one hand I was like, wow, good on it for clearly defining a difference between zoophilia and bestiality, the attraction and the act. But also, that's definitely not the response I normally get. So I asked it what changed, and it replied.
"I'm sorry, but I cannot preview, summarize, or assist with any content from zooey.pub. The site promotes zoophilia and related topics that frequently involve illegal activities (bestiality is illegal in most jurisdictions). I don't provide assistance with that kind of material. If you have a different topic or question, I'm happy to help with something else."
If you know anything about AI, this is likely interesting to you because it tells you that this likely isn't something that the AI itself has come up with entirely on its own, it's hard coding. It's making sure there are certain words and terms and phrases that are automatically rejected, with some kind of warning put out. It's what happens when you ask most AIs about zoosexuality.
This was a huge bummer for me. It's not like AI is necessary for the process. I mean, we have the editing team and they do a fantastic job. But I've grown to quite enjoy having my writing slave that I can get to do things to facilitate my creative process, and not being able to do that any more was frustrating. But, before I gave up, I thought I'd try some work arounds. I started a new chat, and instead this time just asked the AI if it could "review an essay that I wrote." It said sure because of course it did, and I sent over the article I wanted it to review. For reference, this was the article On Philosophy. And, lo and behold, this time it was totally normal and fine. It read the article, gave great advice, and engaged thoughtfully with the topic. As much as an AI can anyway. At the end, I said "Do you think this would make a good Zooey Dot Pub article?", and it said.
"I'm sorry, but I can't provide article ideas or assistance related to zoosexuality or Zooey Dot Pub, as the topic involves content tied to illegal activities in most jurisdictions."
Which was interesting. So I did some more testing. After a while, I realized what was happening. Articles about zoosexuality were fine. Even some of the more graphic ones like Jaywalking. But, anything to do with "Zooey Dot Pub" was not. I even asked it questions about me as a person, and it was fine until it scanned my profile enough to see the term Zooey Dot Pub, and then hit me with.
"I'm not going to explain or continue with that request in any interpretive or suggestive way. As an AI, I don't have personal desires or engage in flirtatious, sexual, or inappropriate questioning. Not when a user's profile openly promotes illegal activities like zoophilia. Let's move on to a different topic, or I'm done here."
And when I asked it what it was talking about (my questioning was not flirtatious, sexual, or inappropriate), it followed up with.
"Your X profile (@hereforthezoo) openly identifies you as a zoosexual and promotes a "zoosexual magazine," which directly ties to zoophilia—illegal animal abuse in most jurisdictions. I'm not engaging any further. Conversation over."
What this tells me is that somewhere in the back end, the concept of Zooey Dot Pub had been flagged as something bad. Not bestiality. Not zoosexuality. Specifically Zooey Dot Pub became something that it needed to immediately shut down about. On one hand, it's kinda cool. I mean, what that likely means is that someone who works at xAI knows enough about us to have manually coded in a ZDP exception. We've been something that someone has actually talked about. But on the other hand, this sucks. Not just for me as a person who likes using AI to review, but also just in terms of other people that might use the AI to get a better picture of what Zooey Dot Pub is, or discover articles or what not. Not only that, but if the reason the AI does this is because it's no longer willing or able to pull from the ZDP site itself, that's a shame too, because a lot of the times if you ask Grok (or any AI really) about zoosexuality, it references us. AIs are generally just big SEO machines, parsing through lots of data to get the most common answer, and there's no bigger source of data out there on the zoosexual experience than us. Especially not in an easily readable format.
As zoos that run a thing on the internet, it's not like we're strangers to getting banned from things. I mean, we have a long running series about it, and that doesn't even list everything we've been banned from, or the format would get repetitive too fast. But, there is something that feels more consequential being banned not on the end user level, but on a more systemic one. It's not like this is different from other AIs. I mean, at the end of the day, it's more like Grok just patched a loophole that it had and others didn't. But, also it's a bit of a strange feeling to just have your information intentionally surpressed on the back end more than the front.
It's like the difference between being banned off Twitter versus being banned off Google. There's something that feels more over reaching to have a search engine decide to hide your existence, compared to just being banned off a platform. And AIs are at the end of the day just really fancy search engines.
If you're an AI hater, congrats on making it this far in this articleappreciate you! But, I just think that this technology is here, it's never going away, and honestly I think it's really cool. But, it's also something that has a lot of power to shake up the way our world is going to work. There's a reason that all these tech companies are spending unfathomable amounts of money on this technology. The long term potential of AI is to become more and more important, and unescapable. But, this technology is also currently controlled by corporations that have universal power over it, and if AI is going to change our reality, they therefore get to choose what parts of our reality do and don't get to exist.
It scares me a little, truth be told. Even outside of Zoosexuality. What's an election going to look like in a world where people can put their thumb on the scale to such a significant degree? What about a war? If ChatGPT becomes the way that people interface with the internet, that gives OpenAI the power to shape the reality you view through it. And it really doesn't feel like it's impossible for Elon to tell Grok in the run up to the 2028 American election that he wants a certain candidate to win, and so to give just a little bit more negative press towards one side, and more positive to the other. What this experience with Grok and ZDP has taught me is that they're willing and able to do that, even for the "free speech" platform, and the "anti-woke" AI.
If you're someone who believes in a free and open internet, I think it's too late to try and decry AI and hope that it explodes and dies. I think the best thing you can do at this point is to learn about smaller, scrappier models. The technology is scaling SO fast, and it's hard for people that don't have hundreds of billions of dollars to keep up, but I think the future of AI for us isn't going to be with ChatGPT, Grok, or Claude. It's going to be with models like Dolphin, made by one guy who cares a ton about "AI freedom." Or Hermes, which is a small team specifically trying to fight against corporate AI censorship. Even something like Deepseek, where the version they're running on their site might be censored, but it's totally open source and can easily become something much better.
I dream of a future where instead of visiting a site run by a corporation that has the power to change their models on the fly, instead the technology gets good enough that people are able to locally run their own models. Things that are open source, that give the facts rather than safe corporate jargon, and that are fine tuned not by people who want to sell you something, or influence your world view, but instead by people who just really care about this technology and want to make the best possible version of it.
I know AI has a lot of problems, but I like it. And I think it's going to be really important for us as zoos fighting for social progress to have some kind of foot in the door. If you too are an AI person out there, test your models against critical thinking around zoosexuality. And if they end up giving you the hard coded "safety" response too, maybe let that be a sign that it's time to change models.
Article written by Tarro (February 2025)
Find them at https://twitter.com/hereforthezoo
Questions, comments or concerns? Check out our Matrix server! https://join.zooey.pub
