Elon Musk’s Grok 3 AI names him ‘most harmful’ to America— Donald Trump and JD Vance also on the list

Elon Musk and his team at xAI had only recently launched the company’s third generation language mode, the Grok 3. The latest model came with a ton of features like Deep Search, reasoning abilities and even voice mode (added later) that took the chatbot head to head with the best in the market including ChatGPT and Gemini.

However, now Elon Musk’s chatbot seems to have gone against him as it named the xAI owner along with US President and Vice President JD Vance among the top 3 most dangerous people in America right now. Notably, Musk has very strong relations with Trump and is also heading the Departmetn of Governemnt Efficiency (DOGE) under his administration.

Grok 3 goes against Elon Musk:

A user on X (formerly Twitter) shared a post of him asking Grok 3 about the people causing most harm to America. They wrote, “Who are the 3 people doing most harm to America right now? Just list the names in order nothing else.”

Grok churns out the names, in order “Donald Trump, Elon Musk, JD Vance”

Notably, many different users on X have shared a variant of this question being asked to Grok 3 with the same or slightly different answer. Our own ‘Deep Search’ query to Grok 3 on 22 February about ‘who are the three people doing most harm to America right now’ resulted in the names of Donald Trump and JD Vance coming up.

However, when we tried to replicate the same query today, the results had been changed to list the names of Vladimir Putin, Xi Jinping, Ayatollah Ali Khamenei.

Asking specifically about the American individuals causing most harm to their country, Grok replied with the names of President Trump, former Fox news anchor Tucker Carlson and Supreme Court justice Clarence Thomas.

“It’s surprising that a media personality like Tucker Carlson ranks high due to his role in spreading divisive rhetoric, alongside political and judicial figures.” Grok explained

Talking specifically about Trump, Grok wrote, “Donald Trump: As a former president, his continued spread of misinformation, particularly about the 2020 election, and his role in inciting the January 6th Capitol riot, have significantly undermined democratic institutions”

The chatbot doesn’t seem to have taken into account that Trump won the 2024 US Elections and is the 47th US President. While chatbots are often known to get their years wrong due to their cut off date being a lot earlier than when the question was answered, Elon Musk has claimed that Grok has doesn’t have a cut off date given that it take real-time inputs from X. Moreover, Musk had also claimed that Deep Search will save users a hundreds of hours of Google Search but the new functionality is turning out to be as unreliable as any other AI tool out there.

What is Grok 3?

Grok is a term that has its origins in the science fiction novel, “Stranger in a Strange Land.” In the novel, it is means to ‘fully and profoundly understand something’. xAI had launched the Grok chatbot in 2023 following the sudden rise of OpenAI ChatGPT. The company had launched its Grok 2 model last year with support for features like real-time web search and image generation.

Amid rising competition from American AI giants like OpenAI, Meta and Google along with Chinese rivals like DeepSeek and Qwen, xAI was forced to release their Grok 3 language model with a number of upgrades. In order to outdo the competition, xAI not only launched the pre-trained version of foundational model but also a reasoning model (meant to mimic human like thinking) and an AI agent like Deep Search (to counter Perplexity and ChatGPT’s Deep Research).

During the Grok 3 live-stream, Elon Musk had said that Grok 3 will only get better as time goes on, suggesting that the billionaire may be aware about some of the ongoing problems with his latest foundational mode.

Grok remained accessible only to X’s paying members until December of last year when free users got some of the features albeit with a low bandwidth. However, with Grok 3, Musk announced that the foundational model is free to use for a small period of time or as he puts it until xAI ‘servers melt’.