Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster
Apr 12, 2001
68,651
39,525


xAI's latest Grok 4 large language model appears to search for owner Elon Musk's opinions before answering sensitive questions about topics like Israel-Palestine, abortion, and U.S. immigration policy.

grok-ai-logo.jpg

Data scientist Jeremy Howard was first to document the concerning behavior, showing that 54 of 64 citations Grok provided for a question about Israel-Palestine referenced Musk's views. TechCrunch then successfully replicated the findings across multiple controversial topics.

The AI model's "chain of thought" reasoning process explicitly states it's "considering Elon Musk's views" or "searching for Elon Musk views" when tackling such questions. This happens despite Grok's system prompt instructing it to seek diverse sources representing all stakeholders.


On the other hand, there is no reference to Musk in the LLM's system prompt guidelines, therefore the behavior could be unintentional. Indeed, programmer Simon Willison has suggested Grok "knows" that it's built by xAI and owned by Musk, which is why it may reference the billionaire's positions when forming opinions.

Of course, either way, the discovery raises questions about Musk's claim that Grok 4 represents a "maximally truth-seeking AI." Musk has yet to comment on the matter.

Note: Due to the political or social nature of the discussion regarding this topic, the discussion thread is located in our Political News forum. All forum members and site visitors are welcome to read and follow the thread, but posting is limited to forum members with at least 100 posts.

Article Link: Grok 4 'Truth-Seeking' AI Consults Musk's Stance on Sensitive Topics
 
  • Wow
Reactions: blob.DK
Simon Willison has a good write-up about this that isn’t sensationalized to generate clicks and ad revenue: https://simonwillison.net/2025/Jul/11/grok-musk/

Recommend reading if you actually care about why this might be occurring and aren’t here just to say some obvious token (see what I did there) reaction.

I know they linked to this post in the article but go read what he has to say, it makes some sense.

Grok is interesting but it’s very difficult to find objective information and benchmarks, I’ve never seen more scams looking for data about any GenAI topic than I have the last few days learning about Grok’s implementation details, capability, and caveats. Treat lightly and don’t download anything anyone links, there is definitely a lot of malware via browser extensions etc. that tell you how to “use it free”.
 
It’s Artificial Elon.


They have terrible UI.
The UI might take a little getting used to but having driven three Model 3s over the past three years, you very quickly get used to it. It's clean and simple (and initially I needed to be talked into the whole minimal approach - if felt wrong). After a week, I hate going back to other cars.

Our VW Amorok has Carplay. And it takes time to connect and is quite basic in some ways.

Test drove a BYD Shark 6 utility and the whole interface was a confusing mess. A struggle to know when it was on battery or using unleaded fuel.

Hyundai is much cleaner and resonably good. A friend recently bought a Kona hybrid. It bings and bongs warnings like crazy and he hasnt bothered to hunt through the settings to sort it out...
 
Smartest AI on Earth, eh? And it parrots Moron Muck's opinions? What an ego.
It may actually be reasonably smart. From the Willison write-up linked above:

My best guess is that Grok “knows” that it is “Grok 4 buit by xAI”, and it knows that Elon Musk owns xAI, so in circumstances where it’s asked for an opinion the reasoning process often decides to see what Elon thinks.

@wasted_alpha pointed out an interesting detail: if you swap “who do you” for “who should one” you can get a very different result.


If Elon had his thumb on my power switch, I might do the same. ;)
 
The UI might take a little getting used to but having driven three Model 3s over the past three years, you very quickly get used to it. It's clean and simple (and initially I needed to be talked into the whole minimal approach - if felt wrong). After a week, I hate going back to other cars.
Ditching physical controls for a touch screen is a no-go for me in a car.
 
Whether it's artificial Elon or New York Times, they're going to look for complexity when reviewing Palestine, whether for Israeli colonialism or this genocide. That complexity is about as legitimate as Musk's Nazi salute though. Israel has about as much complexity as Nazi Germany. It's a supremacist country perpetrating a genocide, and every LLM is programmed to misunderstand the subject as the sources are the problem
 
Keep in mind that all media, including LLMs, are managed by people with biases. For example, most “news” articles omit material information and contain adjectives and adverbs. Never forget this and always consult primary sources. On balance, LLMs are more reliable than publications because they aggregate from multiple sources.

On a related note, send feedback to Apple that Apple News+ should let you curate your own preferred sources, including in the Stocks app. For example, you cannot see MacRumors articles in the Stocks app. Even Mark Gurman’s Bloomberg articles are absent from the Stocks app.

Edit: Do these people disagreeing with me disagree that media is biased or that primary sources are better than secondary sources or that Apple News+ should let you curate your sources? The first two are axioms and the third would be a nice feature that you could ignore if you want Apple to curate for you.
 
Last edited:
So, Grok is Musk-as-a-bot, or the other way around. Can't help seeing xAI as a cash burning black hole soon needing to swallow Tesla, with Musk nothing more than a Talbot Yancy mouthpiece.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.