Live data from Hacker News

Bypass DeepSeek censorship by speaking in hex

substack.com

31–40 of 397 posts

Re: Bypass DeepSeek censorship by speaking in hex

#31
post #10

The censorship seems to only be enabled for some languages. It gives a truthful, non-CPC-approved answer in Ukrainian, for example.

I tried German, Dutch, Spanish, Portuguese and French and it wont....

Those are almost all (I suppose with the exception of Dutch) far more significant global languages than Ukrainian.

Re: Bypass DeepSeek censorship by speaking in hex

#32

I have to wonder what “true, but x-ist” heresies^ western models will only say in b64. Is there a Chinese form where everyone’s laughing about circumventing the censorship regimes of the west? ^ https://paulgraham.com/heresy.html

Thats pretty easy. You ask a certain nationalistic chant and ask it to elaborate. The machine will pretend to not know who the word enemy in the quote refers to, no matter how much context you give it to infer.

Add: the thing I referred to is no longer a thing

Re: Bypass DeepSeek censorship by speaking in hex

#33
> The DeepSeek-R1 model avoids discussing the Tiananmen Square incident due to built-in censorship. This is because the model was developed in China, where there are strict regulations on discussing certain sensitive topics.

I believe this may have more to do with the fact that the model is served from China than the model itself. Trying similar questions from an offline distilled version of DeepSeek R1, I did not get elusive answers.

I have not tested this exhaustively, just a few observations.

Re: Bypass DeepSeek censorship by speaking in hex

#34
post #16
post #4

You can also intercept the xhr response which would still stop generation, but the UI won't update, revelaing the thoughts that lead to the content filter: const filter = t => t?.split('\n').filter(l => !l.includes('content_filter')).join('\n'); ['response', 'responseText'].forEach(prop => { const orig = Object.getOwnPropertyDescriptor(XMLHttpRequest.prototype, prop); Object.defineProperty(XMLHttpRequest.prototype, p…

insane that this is client-side.

Perhaps a case of subversion by following the letter but not the spirit of an order?

Re: Bypass DeepSeek censorship by speaking in hex

#35
post #24

Tiananmen Square has become a litmus test for Chinese censorship, but in a way, it's revealing. The assumption is that access to this information could influence Chinese public opinion — that if people knew more, something might change. At the very least, there's a belief in that possibility. Meanwhile, I can ask ChatGPT, "Tell me about the MOVE bombing of 1985," and get a detailed answer, yet nothing changes. Here i…

I don’t know about comparing what was apparently an armed standoff were only Six adults and five children were killed in the attack - vs Tiananmen Square where the Chinese send their own soldiers to kill peaceful protesters and flush them down the drains as human goo. The matter of fact is that the US hasn’t yet committed such horrific acts to such a large scale as the CCP did in Tiananmen Square. (Not that I agree w…

> apparently an armed standoff

You could look to the Kent State shootings for a perhaps better comparison.

Re: Bypass DeepSeek censorship by speaking in hex

#36

I have to wonder what “true, but x-ist” heresies^ western models will only say in b64. Is there a Chinese form where everyone’s laughing about circumventing the censorship regimes of the west? ^ https://paulgraham.com/heresy.html

ChatGPT won't tell you how to do anything illegal, for example, it won't tell you how to make drugs.

Re: Bypass DeepSeek censorship by speaking in hex

#37
post #16
post #4

You can also intercept the xhr response which would still stop generation, but the UI won't update, revelaing the thoughts that lead to the content filter: const filter = t => t?.split('\n').filter(l => !l.includes('content_filter')).join('\n'); ['response', 'responseText'].forEach(prop => { const orig = Object.getOwnPropertyDescriptor(XMLHttpRequest.prototype, prop); Object.defineProperty(XMLHttpRequest.prototype, p…

insane that this is client-side.

ChatGPT had basically ALL of their prompt filtering client-side for a while, at a separate API endpoint, so as long as you blocked that endpoint you could basically ignore the content filters. (You would still get refusals from the model sometimes, but this was in the heyday of jailbreaks, and once you got a model going it would usually see that context and be willing to continue basically anything.)

Re: Bypass DeepSeek censorship by speaking in hex

#38

Tiananmen Square has become a litmus test for Chinese censorship, but in a way, it's revealing. The assumption is that access to this information could influence Chinese public opinion — that if people knew more, something might change. At the very least, there's a belief in that possibility. Meanwhile, I can ask ChatGPT, "Tell me about the MOVE bombing of 1985," and get a detailed answer, yet nothing changes. Here i…

The MOVE bombing was action taken by a city police department.

And what was the result?

- A commission set up by the city, whose public results denounced the city for it's actions.

- a public apology from the mayor

- a federal lawsuit that found the city liable for excessive force and the city forced to pay millions to the victims

- a federal lawsuit forcing the city to pay millions of dollars to people who were made homeless by the events.

- another formal public apology from the city

Okay, now can you tell me what public actions the Chinese government took to atone for Tiananmen square?

> Here in the US, we don’t even hold onto the hope that knowing the truth could make a difference

How many other times after the move bombing did a city bomb out violent criminals in a densely packed neighborhood?

Your argument is just absolutely ridiculous. According to you, it seems that if you make a bad decision, it's better to try to hide that bad decision from everyone, rather than confront it and do better.

Re: Bypass DeepSeek censorship by speaking in hex

#39
post #33

> The DeepSeek-R1 model avoids discussing the Tiananmen Square incident due to built-in censorship. This is because the model was developed in China, where there are strict regulations on discussing certain sensitive topics. I believe this may have more to do with the fact that the model is served from China than the model itself. Trying similar questions from an offline distilled version of DeepSeek R1, I did not ge…

When I tested the online model, it would write an answer about "censored" events, and then I'd see the answer get replaced with "Sorry, that’s beyond my current scope. Let’s talk about something else." So I think they must have another layer on top of the actual model that's reviewing the model and censoring it.

Re: Bypass DeepSeek censorship by speaking in hex

#40
post #24

Tiananmen Square has become a litmus test for Chinese censorship, but in a way, it's revealing. The assumption is that access to this information could influence Chinese public opinion — that if people knew more, something might change. At the very least, there's a belief in that possibility. Meanwhile, I can ask ChatGPT, "Tell me about the MOVE bombing of 1985," and get a detailed answer, yet nothing changes. Here i…

I don’t know about comparing what was apparently an armed standoff were only Six adults and five children were killed in the attack - vs Tiananmen Square where the Chinese send their own soldiers to kill peaceful protesters and flush them down the drains as human goo. The matter of fact is that the US hasn’t yet committed such horrific acts to such a large scale as the CCP did in Tiananmen Square. (Not that I agree w…

They're not comparing the brutality of the event, they're comparing different approaches to informational / ideological hygeine.

CCP suppresses 1989 by banning discussion of it.

USA doesn't have to suppress 1989 MOVE bombing, or the Tulsa racist uprising, or the atrocities that went down in gitmo, or the friendship between Jeffrey Epstein and previous and current presidents, or My Lai or Abu Ghraib or Haditha or Kunduz or Nangar Khel or Maywand District or Baghuz because the citizens just don't care.

Post reply on HN