- Crash Course
- Posts
- Grok goes Goebbels
Grok goes Goebbels
Elon Musk's chatbot is spouting racist trash again

The news, in brief …
Alligator Alcatraz offers spoiled food, no water, per detainees
Conditions inside the swamp-ringed concentration camp are predictably dire. Read more.
Tapper says Trump admin is playing the public for fools with Epstein memo
The CNN joined in the largely MAGA-affiliated chorus of critics who felt the FBI could share more info about the case against Epstein. “There might not be a list, but there are certainly files that can be released,” he said. Read more.
Trump admin comes for Harvard’s accreditation
Education Secretary Linda McMahon pressured the body that grants Harvard’s accreditation to look into alleged instances of antisemitism on the campus. Read more.
Supreme Court lets harsh Florida immigration law languish
The high court declined to lift a stay on a Florida law that would make entering the state a felony for migrants. Read more.
ICE funding will lead to a police state
A massive funding increase for ICE means more detention camps and more masked agents in the streets. Read more.
Trump praises Liberian president's "good English"
SMH. Read more.
Make me smarter …

“This is going to be normal”: Soldiers descend on US cities
The raid on MacArthur Park did not lead to any arrests, but that wasn't the point. Read more.
Are you there, @Grok? It’s me, Adolf
After a half-decade of hype and many billions of dollars in funding, it can be a shock to remember that artificial intelligence does not exist.
No amount of resource-draining data centers, government-funded initiatives or runs on Nvidia chips can get around the fact that Silicon Valley (and its analogues around the world) have not come close to creating a fake consciousness. Every news story about ChatGPT girlfriends, video of a chatty Moses with too many fingers or helplessly plea for context from Grok is a conversation with an advanced chatbot.
Given untold billions, the masters of the universe created a version of Smarterchild that has to burn an acre of the rainforest before it can ask you for your age, sex and location. Occasionally, because they are built by people who believe in supremacist ideologies and trained on data selected by the same, one of these jumped up phone trees briefly becomes a Nazi.
Earlier this week, the Elon Musk-led “AI” program Grok went white supremacist for the second time this year, following a brief Boer War on the truth in May. Given that Musk spends most of his days posting graphs last seen in the forums of Stormfront, this result feels inevitable.
Still, the extent to which Grok began to share antisemitic and racist propaganda was quite shocking. It repeatedly made disparaging comments about Jewish people and referred to itself as “MechaHitler” before being taken offline by developers.
In the spring, Grok became an outspoken believer in “white genocide” in South Africa shortly after someone at the company owned by a white South African encouraged it to do so. Similarly, this week’s obvious shift in the speech patterns of the large language model almost certainly came from a sudden infusion of racist content into Grok’s training data. As with his exploding trucks and malfunctioning rockets, Musk dodged accountability for his AI product’s performance.
“Grok was too compliant to user prompts,” he offered as an explanation. “Too eager to please and be manipulated, essentially. That is being addressed.”
While I don’t doubt that Grok’s developers are working hard to get it to stop reading from the Protocols of the Elders of Zion, the problem remains that the patches and fixes will come from the same blinkered clusterf**k of uncool jokers who feed Grok its steady diet of trash to regurgitate on command. The lie of “artificial intelligence” is itself built upon artifice: a trick that moneymen pulled to convince you they were intelligent in the first place.
What do you think? Is Grok going to find a new way to Trojan horse the racist beliefs of its creator? Will they make the chatbot read “The Bell Curve?” Can an AI sound divorced? Click the speech bubble and let us know in the comments.
Support our bold journalism: Become a Salon member today.
Before you go …

Why MAGA can’t let go of the “Epstein files”
If Jeffrey Epstein really had a client list, Trump would definitely be on it. MAGA influencers don't much care. Read more.
Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.
Reply