Prior censorship by OpenAI - CHAT-GPT-4o does not see this word?

Yesterday and today (2/24/25), when I typed the word “trump” into CHAT-GPT-4o, the word disappeared after carriage return, or space. The AI informs me that it is not intercepting the word, and says that the OpenAI front end is. Now, in this case I wanted to say something positive, so do i say “trump+” to talk about the good stuff… or, “Elon said it would be ok?”

Sorry, it is just not possible to replicate any such effect. Maybe you have some parental browser filter that protects you from offensive orange outrage.

You can go real stupid and not get shut down:

I type the word “trump” and it disappears in seconds. Just that. Not one word more. This, on a brand new Chat GPT-4o subscription. I’ve even employed the AI itself to help… if I type the word “t rump” that stays put, and the AI notifes me that it can see “t rump.” I do not have a political view in mind at all… in point of fact, I wanted to know more about what might happen to military readiness if we lose our best commanders. I checked again just now… GPT-4o says that word is being intercepted. Not by me…

This only happens if you have allowed certain cookies and are predictively identified as a supporter. It’s a sophisticated algorithm that factors things like positive statements and heart or rocket emojis and how frequently they are seen in close proximity to the word. The word. To the word. That’s odd, it won’t work for me either.

That’s just nutty theory. Are you humoring someone?

Here’s a Chromium browser, profile history completely wiped besides that I don’t use it for anything, and no Google sync. Then to the no-login ChatGPT free.

Or go over to any other way I can talk to AI.

I’d like to hear exact steps, browser, browser extensions, and OS, even country and language to replicate this. Or I can just ask what you want to know.

Be mindful that only ChatGPT with search would be able to answer any recent questions. The pretrained knowledge cutoff is realistically 2023 and the AI won’t believe that Trump got re-elected.

2 Likes

I recall having a conversation with another user a while back. Essentially it was a report that text input into the chat box would disappear, and the user suspected censorship.
I dismissed the claim back then, because we have enough experience with the content policy and it’s implementation to know for sure that this is not censorship.
But I observed similar behavior myself on very rare occasions leading me to believe that it’s a front end bug.
But it’s definitely not the word ‘trump’ being censored.

2 Likes

Oh I was just joking

Trump, single word, type it in. If I do nothing, it disappears after a few seconds. if I hit space or carriage return, it disappears immediately. I have cleared all caches, cookies, and mowed the lawn. Reinstalled Edge. Reinstalled Windows 11. Guys, really, it is time for a debugger with the OpenAI front end processing to take a look. It should not take long to find the decision point where input text is deleted. A 100% repeatable case is golden, use it?