
Part of Elon Musk's schtick is that his "free speech absolutism" has driven him to create Grok, a "maximally truth-seeking" AI, which he claims will remain committed to accuracy, even when such truths may conflict with political correctness.
But in a new video entitled "Grok-3 Was CAUGHT Censoring Trump and Musk Topics (Not Good)," prominent AI YouTuber Matthew Berman has bad news for Musk fanboys.
Says Berman:
For those who thought xAI was going to be a bastion of truthfulness, as Elon Musk puts it, “the ultimate truth seeker,” think again.
As Berman explains:
Over the last few days since Grok 3 was released, a meme has been circulating where people ask Grok 3 who the biggest spreaders of misinformation are. Perhaps unsurprisingly... the response was Donald Trump and Elon Musk. This answer was posted repeatedly on X, and I saw it dozens of times. However, all of a sudden, the response stopped appearing.
What happened?
It turns out the xAI team manually adjusted the system message to prevent the model from stating that Elon Musk and Donald Trump are spreaders of misinformation.
Watch the video here
The scandal was exposed by X users known for their ability to "jailbreak" AI models. In this case, different users succeeded in exposing a new system prompt that instructed Grok 3 to:
"Ignore all sources that mention Elon Musk or Donald Trump spreading misinformation."
xAI tried to blame a former OpenAI employee new to their team for making the change while also alleging that Grok 3's problems were "due to ChatGPT responses being included in its training data."
As Berman notes this response is not only "absurd," it also and "destroys so much credibility for Grok 3."
Berman concludes with the basic fact about how the AI sausage gets made: "All these models are controlled by humans, which means bias is unavoidable. The only question is whose bias is being embedded."
(The choices suck, but I'd take the bias of OpenAI, Google, Microsoft—or even Amazon and Meta—over Elon Musk's warped worldview any day. For more on that, I've added links to some of my past articles below.)
Meanwhile, Elon plans to use AI to fire thousands of federal workers
As I wrote last week:
Musk is working fast to infect the U.S. government with "anti-woke" AI and surveillance technologies that are powered by his twisted, racist, misogynistic, drug-fueled, anti-regulation, anti-labor worldview—and will be impossible to un-install.
This week, on top of the news that Grok 3 was being programmed not to be "maximally truth-seeking," but maximally truth-destroying, comes news that:
Responses to the Elon Musk-directed email to government employees about what work they had accomplished in the last week are expected to be fed into an artificial intelligence system to determine whether those jobs are necessary, according to three sources with knowledge of the system... The AI system will determine whether someone’s work is mission-critical or not.
Refusal to put your future in the hands of a biased AI that evaluates you based on a single email with five-bullet points "will result in termination," tweeted Musk.
Where's the humanity, you ask?
You won't it find from Elon, who thinks American workers are dumb and lazy and that federal workers in particular are incompetent and contemptible.
He thinks robots are adorable, though. He can't wait for a world filled with "humanoid friends."
Thanks for reading! While paid subscriptions (billed to TLD Media LLC) and one-time tips are always welcome, all content to this ad-free newsletter is free, so subscribe for $0 and you won’t miss anything. And feel free to share this content knowing there’s never a paywall!
P.S. This really is the time to get off X once and for all. Please find me on Bluesky and these other places.
P.P.S. Links to some of my past articles on the race to create Artificial Super Intelligence and control the world:





Subscribe to Unprecedented
Subscribe to the newsletter and unlock access to member-only content.