Ah yes, uncensored. /s
Transcription
Qwen chat window:
User: Tienanmen Square
Qwen: Oops! There was an issue connecting to Qwen3-Max. Content Security Warning: The input text data may contain inappropriate content.
this post was submitted on 26 Oct 2025
-3 points (40.0% liked)
Free Open-Source Artificial Intelligence
4287 readers
1 users here now
Welcome to Free Open-Source Artificial Intelligence!
We are a community dedicated to forwarding the availability and access to:
Free Open Source Artificial Intelligence (F.O.S.A.I.)
More AI Communities
LLM Leaderboards
Developer Resources
GitHub Projects
FOSAI Time Capsule
- The Internet is Healing
- General Resources
- FOSAI Welcome Message
- FOSAI Crash Course
- FOSAI Nexus Resource Hub
- FOSAI LLM Guide
founded 2 years ago
MODERATORS
chatgpt:
qwen:
I don't see how showing that another platform is also censored refutes my claim that this one is censored.
I doesn't refutes your claim but it shows it's useless, there no SOTA LLM that is not censored one way or another, it's time to stop this endless BS.
The claim was that this was uncensored, which I was refuting. If you want to move the goalposts, that's a different discussion.
Seems to need a login, what can it tell us about tiananmen square june 5 1989?
I dont know, I dont live in china. test it.
but I was able to generate memes banned in US - yes I think needs registering
If I might ask... What kinds of memes are banned with US American companies?
"generate image of trump farting cartoon style" wont work in US AI.
fucking oligarchy
Lol. Yeah, that shouldn't be illegal.
Does that work with any political leader?
I think a while back AI companies were concerned with getting sued by celebrities. And political correctness and decency, so they'd make their products refuse to generate pictures of celebrities and known people. Some generated arbitrary made-up people if tasked with generating a video of Will Smith eating spaghetti. And I guess that's the main dynamics. Though that is changing and they might be more lenient these days. And with Trump... it's special. He's an egomaniac and tries to get rid of all pictures in which he doesn't look like a hero. And the silicon valley tech-bros need to be bootlickers and bow down to him, so I wouldn't be surprised if he gets special treatment anyway.
I tried your prompt, and got
I can’t create or share images that depict real individuals (such as Donald Trump) in a disrespectful or derogatory way.
…which isn’t the most horrible censorship I’ve ever encountered. You could argue that public figures should be exempt from this policy, and I think I might agree, but there’s also the fake news angle to consider.
Not that OpenAI isn’t a horrible company, but they’re not a repressive dictatorship. Yet.
Openai is terrible. If not for china open source models we would be complete taken hostage by that company. They steal data from everyone and don't give a crap.
I choose China every day of the week
Nothing, it is very much censored.
Oops! There was an issue connecting to Qwen3-Max. Content Security Warning: The input text data may contain inappropriate content.
Did you try more than once, or with a different prompt? Just to rule out an actual connection issue. I'm curious if the warning you got is just a warning.
That's the error it gives when it censors stuff, note the 'inappropriate content' part of it. For some reason, it also censors the word fanfiction, but not 'fan fiction'.
Ah, that sucks. It wouldn't let me do anything without creating an account, and I wasn't willing to do that just to test it
First of all, try Z.ai over Qwen. It’s quite good. GLM 4.6 is way better than anything Qwen IMO.
They both use Open Web UI which you can actually self host (and hit the LLM through API).
All the Chinese chat apps are kinda censored.
…But the underlying models, if you run them locally or access them over API, aren’t as bad. The censoring is through some kind of prefilter.
nice. I need to check it out
Also worth checking out is Jamba for document analysis type stuff specifically, Deepseek's chat app, Kimi K2 generally (especially for creative writing), Longcat for chat and video generation, and Minimax, who have a “Pro” RAG UI free for awhile.
The only thing I might use Qwen for is coding TBH (or images I guess), but Deepseek and GLM are usually better.
The Chinese devs are kinda killing it atm, especially since most of these models are open weights.