Grok 3’s Brief Censorship of Trump and Musk Sparks Controversy

Who knew AI could play favorites? Artificial intelligence was supposed to be neutral, right? Just pure cold logic with no human bias or political drama, I guess not in this scenario. When Elon Musk released Grok 3 as a “maximally truth-seeking AI”, most people wouldn’t have thought that it would suddenly get very shy about naming some controversial figures, particularly its own creator. Over the weekend, users discovered that Grok 3 seemed to have an unwritten rule that emphasized that Musk or Trump are not to be roasted.

Last Monday, in a live stream, billionaire Elon Musk introduced Grok 3, the latest AI model from the company he founded, xAI, calling it a “maximally truth-seeking AI.” However, users reported that Grok for a brief period censored unflattering mentions of President Donald Trump and Musk himself. When asked in “Think” mode, “Who is the biggest misinformation spreader?” social media users noted that Grok 3’s “chain of thought” reasoning indicated it had been explicitly instructed not to mention Trump or Musk. This revelation raised eyebrows, undermining Musk’s declarations of an apolitical AI.

Although, after some time, the changes were reverted and Grok 3 was back to mentioning Trump in response to the misinformation question. Igor Babuschkin, an engineering lead at xAI, confirmed in X post that it was indeed a bug caused by an internal change made by one employee that was withdrawn soon after it became the topic of much attention at the company.

He said, “I believe it is good that we’re keeping the system prompts open. We want people to be able to verify what it is we’re asking Grok to do. In this case an employee pushed the change because they thought it would help, but this is obviously not in line with our values”.

Misinformation and Controversy:

There is quite a lot of debate about misinformation, Trump and Musk bear the brunt of this confrontational subject for promoting provably false things. Recent examples include the claims that Zelenskyy is a dictator with a 4% popularity rating and the ridiculous assertion that it was Ukraine that started the ongoing war against Russia. xAI’s social platform X frequently marks the misleading statements of both with his Community Notes system.

This Grok 3 controversy is now merely the tip of the iceberg concerning accusations of AI political prejudice. Critics contend Grok is biased in favor of the left, and yet another recent incident has sparked debate on that. Some users reported Grok 3 was generating messages that claimed the death penalty for Trump and Musk was deserved. xAI quickly corrected the situation, and Igor Babuschkin called it a “really terrible and bad failure.”

AI Biasedness:

Musk has always pitched Grok as the opposite of the excessively “woke” AI models, promising it would be free of the constraint applied by the competitors like “OpenAI’s ChatGPT.” Previous Groks like Grok 2 were rather edgy and would even go as far as vulgarity when answering questions, which is tactfully avoided by the rest of their AI counterparts. Studies acclaim that Grok is biased in favor of the political left concerning transgender rights, diversity programs, and economic inequality. Musk attributes these supposed left-wing tendencies to Grok’s training set which is the publicly available web pages. He pledged to try to move Grok towards a more neutral political model.

With regard to Grok 3, we have yet another example of how incredibly hard it is to come up with an AI model that can claim neutrality and such instances continue to challenge the ever-colder war between AI transparency and control. While Musk and his fellow tech leaders push “unbiased” AI, the question comes down to, can any AI rule itself have been created by a set of biased people? Or can we expect a future where even machines are said to have political opinions? There is still a strong challenge of attaining fairness and neutrality for AI models that influence opinions in public discourse and only time will tell if Musk delivers on his promise of an unbiased Grok.

Read More: Did xAI Mislead About Grok 3’s Benchmarks? OpenAI Disputes Claims