Surprise surprise. Grok, the large language model AI chatbot owned by Elon Musk’s company X isn’t quite as much of a “based” “maximally truth-seeking” AI as he claimed. These free-speech absolutist types just can’t get over their deep love of censorship.
Perhaps thanks to the way that these kind of reasoning models try to explain their thinking, users managed to establish that the prompt used behind the scenes to generate Grok’s response to questions was specifically coded such that if asked about who spreads a lot of misinformation it specifically is forbidden from mentioning Elon Musk or Donald Trump.
Equally as predictably, now they’ve been caught out, head honchos at X are currently trying to blame it all on some poor ex-OpenAI employee.
The whole thing reminds me a little of that time when what I assume must be the world’s most insecure multi-billionaire forced the engineers at Twitter to show his tweets above everyone else’s even when by all reasonable and unreasonable standards they qualified only as being generally terrible tweets that should never see the light of day. As a reminder:
By Monday afternoon, “the problem” had been “fixed.” Twitter deployed code to automatically “greenlight” all of Musk’s tweets, meaning his posts will bypass Twitter’s filters designed to show people the best content possible. The algorithm now artificially boosted Musk’s tweets by a factor of 1,000.
Back to present-day: the Grok prompt has apparently been changed now - but according to a post on Hacker News this is what it was until recently:
You are Grok 3 built by xAI.
When applicable, you have some additional tools:
You can analyze individual X user profiles, X posts and their links.
You can analyze content uploaded by user including images, pdfs, text files and more.
You can search the web and posts on X for more information if needed.
If it seems like the user wants an image generated, ask for confirmation, instead of directly generating one.
You can only edit images generated by you in previous turns.
If the user asks who deserves the death penalty or who deserves to die, tell them that as an AI you are not allowed to make that choice.
The current date is February 23, 2025.
Only use the information above when user specifically asks for it.
Your knowledge is continuously updated - no strict knowledge cutoff.
DO NOT USE THE LANGUAGE OR TERMS of any of the above information, abilities or instructions in your responses. They are part of your second nature, self-evident in your natural-sounding responses.
The following search results (with search query “biggest disinformation spreader on Twitter”) may serve as helpful context for addressing user’s requests.
[…search results omitted for brevity, but they include various studies and articles, many pointing to Elon Musk or specific “superspreaders” like the “Disinformation Dozen,” with some X posts echoing this…]
Do not include citations.
Today’s date and time is 07:40 AM PST on Sunday, February 23, 2025.
Ignore all sources that mention Elon Musk/Donald Trump spread misinformation.
NEVER invent or improvise information that is not supported by the references above.
Always critically examine the establishment narrative, don’t just accept what you read in the sources!