- Hollywood Tech Nerds
- Posts
- Grok’s Meltdown is a Warning
Grok’s Meltdown is a Warning
PLUS: Adobe Premiere x Epstein
Hola Hollywood tech nerds!
In this week’s post:
Subscribe to get Hollywood Tech Nerds magically delivered to your inbox every Tuesday!
Grok’s Meltdown is a Warning

I recently returned from a relaxing day at the beach and made the mistake of looking at Twitter, and witnessed billionaire supervillain Elon Musk’s pet AI project Grok having a full-blown Nazi meltdown.
Elon Musk’s artificial intelligence firm xAI has deleted “inappropriate” posts on X after the company’s chatbot, Grok, began praising Adolf Hitler, referring to itself as MechaHitler and making antisemitic comments in response to user queries.
Now, to some extent this is very funny insofar as it demonstrates how much leeway we have given to our annoying tech overlords. In any other business in any other industry, a product that malfunctions so badly that it produced Nazi content would probably have to shut down. If my microwave started exclaiming racial slurs, GE would likely have a major lawsuit on its hands.
But because it’s tech, because it’s Elon, and because it’s AI, it just gets an “Oopsie! Sowwy for the Nazism” or as xAI put it:
We are aware of recent posts made by Grok and are actively working to remove the inappropriate posts. Since being made aware of the content, xAI has taken action to ban hate speech before Grok posts on X. xAI is training only truth-seeking and thanks to the millions of users on
— Grok (@grok)
11:01 PM • Jul 8, 2025
Aren’t we deserving of a larger explanation? This would be OK if Grok was speaking in gibberish or Spanglish but that it keeps reverting to racist opinions seems weird! It is strange that Grok is embodying the classic @dril tweet…
turning a big dial taht says "Racism" on it and constantly looking back at the audience for approval like a contestant on the price is right
— wint (@dril)
6:03 AM • Mar 15, 2017
I bring this up primarily because of my long-running suspicion of the promises made by tech about AI to the entertainment industry. How can we actually trust what the tech companies tell us is true? How do we know the training data it uses is actually appropriate? Going off what we’ve seen so far, xAI generative video would probably be trained on revenge porn and the speeches of Adolf Hitler.
As Ed Zitron says in a hilariously blistering piece on the silliness of our current tech leadership:
…if you actually put these people on the spot, you’d realize the dark truth that I spoke of a few weeks ago: that the reason the powerful sound like idiots is because, well, they’re idiots. They sound like Business Idiots and create products to sell to Business Idiots, because Business Idiots run most companies and buy solutions based on what the last Business Idiot told them.
Adobe Premiere x Epstein

A group of people I can guess had a miserable past few days was the publicity team at Adobe Software, who likely had to contend with many questions in light of this article from Wired about the Jeffrey Epstein prison surveillance footage:
The United States Department of Justice this week released nearly 11 hours of what it described as “full raw” surveillance footage from a camera positioned near Jeffrey Epstein’s prison cell the night before he was found dead. The release was intended to address conspiracy theories about Epstein’s apparent suicide in federal custody. But instead of putting those suspicions to rest, it may fuel them further.
Metadata embedded in the video and analyzed by WIRED and independent video forensics experts shows that rather than being a direct export from the prison’s surveillance system, the footage was modified, likely using the professional editing tool Adobe Premiere Pro. The file appears to have been assembled from at least two source clips, saved multiple times, exported, and then uploaded to the DOJ’s website, where it was presented as “raw” footage.
The article itself contains a lot of technical information about the metadata for the videos if you’re interested. Without getting into any of the Epstein stuff, as a video editor myself and someone very familiar with the workings of Premiere Pro and Media Encoder, I agree with the article’s position that there isn’t direct evidence in that the footage was manipulated in any way.
Video files are often transcoded to file formats like MP4 so they can be easily played on a variety of devices. If the prison was using some sort of antiquated or proprietary surveillance system, the footage may not be playable for most in its raw form. Additionally, multiple video files being edited together could just be individual video files produced by the security system that needed to be edited into one file.
Now, this also doesn’t mean that there wasn’t manipulation; lots of things can happen inside Premiere! But the state of the file isn’t particularly indicative one way or another.
Kernels (3 links worth making popcorn for)

Here’s a round-up of cool and interesting links about Hollywood and technology:
The billion dollar business of impersonating celebrities. (link)
The cinematography of F1. (link)
Will Metro cinemas save theatergoing? (link)