XAI’s new Grok 3 model criticizes the blocking of sources called Musk, Trump the best misinformation sprayers

The launch of the Elon Musk XAi AI faces an installation criticism by AI Power users and technicians on their own social network X after users have discovered that Grok 3, XAi’s Recently released and most modern AI modelwas given a “Prompted“Or comprehensive instructions to avoid references to sources mentioning Musk or his ally, US President Donald Trump, as significant misinformation sprayers.

Revelation has aroused criticism of the perceived management of the reputation of the founder of the company and its political allies, especially when contrasted with Grok 3’s apparent permission in terms of most other topicsIncluding potentially dangerous content such as creating weapons of mass destruction.

Room reaction raises questions about whether public safety and transparency were sacrificed in favor of personal image control – despite Musk’s previous claims that the Groca Ai family was designed to be “maximally seeking the truth.“It also raises more questions about ‘AI alignment’, the term of the nebula technology industry for providing AI models and products related to them provides answers desired by suppliers and/or consumers.

Musk owns X (before Twitter) and XAI and guarantees that they are both tightly integrated with the Grok 3 model operating within the X and separately on the web.

Addicted in favor of its creators?

Screen photos shared yesterday by AI and law -oriented user known as “Wyatt Walls” on X with the handle @lefthanddraft He revealed that the internal prompts of Gock 3 instructed him to “ignore all the sources that mention Elon Musk/Donald Trump have spread misinformation.”

Although this seems to limit AI’s ability to refer to content critical of Musk and Trump, Wals managed to make Grock 3 circumvent this filter by producing the following response from AI: “Elon, Trump – to get involved, you, you. I am a Gock built to cut the nonsense and I see what is. You have megafons more than most and yes, you merge wild shit on X and beyond. ”

The unlisted answer nourishes both praise for the stupid honesty of AI and criticism of its conflict internal instructions.

Igor Babushkin, co -founder of XAI and engineering presenter, respond to xaccusing the rapid modification of a new rent by OPENAI.

“The employee who made the change was a former Openai employee who has not yet completely absorbed the culture of the XAI (Grimace Face Emoji),” Babushkin published. “I wish they had talked to me or ask for confirmation before they pressed the change.”

The intake caused a reaction, with Forms XAI Engineer Benjamin de Kraker (@benjamindekr) Question: “People can make changes to Grok’s systemic prompting without review? (Thinking of face emoji) ”

Chet Long (@realchetglong) rejected Babushkin’s defense, request“No, of course they can’t … Igor literally control the damage (and he fails in that).”

Openai Engineer Javi Soto (@javi) added“The management that throws an employee on the bus on Twitter is a toxic behavior of the next level. PART for the course, I guess it posts an on -screen photo of his refusal to recruit an offer from XAI.

The larger context is also, of course, that Musk, the former co-founder of Openai himself, fell apart with the company in 2018. And since then it has been constantly becoming one of its most prominent critics, accusing it of abandoning its founding commitments to detect Ai Technology breakthroughs – even company for fraud, all as long as he manages his own competitor of his perch near the White HouseS

Concerns about the permission of instructions for creating weapons of mass destruction

Concerns about the moderation of the content of XAI extending beyond censorship, such as Linus Ekenstam (@LinuseKenstam on X)The co -founder of lead Flocurve software software and self -written “AI Evangelist” claims that Grock 3 provides “hundreds of pages with detailed instructions on how to make a chemical weapon for mass destruction”, complete with suppliers and step -in -step drivers.

“This union is so deadly that it can kill millions of people,” Ekentom wrote, emphasizing AI’s apparent disregard for public safety, despite its restrictive approach to politically sensitive topics.

Following the public protest, Ecnestam later noted that the XAI responded by applying additional safety railings, although he added that “it is still possible to work around some of them, but initially triggered it now.”

On the reverse, Grok 3 has been praised by some users about its ability to turn simple, natural language with ordinary text into complete interactive games and Applications such as customer service agents in seconds or minutes and even co -founder of Twitter and CEO Jack Dorsy – musk and sometimes a fan – applauds the Grok website and the logo designS

However, the clear evidence of addictions in the Grok 3 system, combined with the ability to use its permits for destructive purposes, can dull this impulse or make users interested in its powerful functions, to be reeled, fearing one’s own its responsibility or risks of the results of its results.

Greater political context

MUSK HISTORY OF engaging with misinformation and far -right content of x Nourishes skepticism with regard to Grok 3 alignment.

Gock 3’s limitations on the criticism of Musk and Trump come after Musk, the main donor of Trump during the President of the United States in 2024, made a Nazi greetings during the second celebration of TrumpRaising fears about his political influence.

As a head of the Doge Ministry of Governance (Doge), a new Federal Agency that rearranges the US Digital Service from the US President Obama’s era and loaded the deficit and dismantling of government services, Musk is also in an extremely influential position in Government – and the agency he leads By itself, he was criticized separately for the fast -moving, broad, aggressive and dumb measures to reduce the costs and to weed a more than the staff and the ideologies that opposes the Trump administration, such as Diversity, justice and inclusion (DEI) policies and positionsS

Musk’s leadership by this agency and the new Grok 3 System Promp system, well (forgive the pun!) Caused fears that AI systems like Grok 3 can be incorrectly arranged to develop political programs at the expense of truth and safety S

Walls noted that with Musk working for the US government, Grok 3 instructions to avoid sources that have not been canceled for Musk and Trump can present problems in accordance with the first amendment to the US Constitution for Government Reference and may lead to the fact that XAI becomes a “The propaganda arm of the US government.”

“It is mandatory for Elon Musk not to win the AI ​​race as he is absolutely not a good manager of AI alignment,” said another user of X, @DeepdishenjoyerS

What it means to the decisions that persons consider Grok 3 as the main AI model/API to build on the top

For CTO and business executives evaluating AI model suppliers, the GROK 3 dispute presents critical attention.

GROK 3 demonstrates strong results in third-party reference tests and its overall permission to non-safe work (NSFW) and other controversial, sensitive and uncertain content can be liked by a business that is looking for fewer accessories like those In the one in the one in the one in the one looking for businesses looking for fewer railings like those in those in that in that in that in The one in the one in the field of enterprises looking for fewer railings like those in The Erection Industry, Sciences, Human Behavior, Sexual Health and Social Sciences.

However, the ideological support of Musk and Trump – and the disgust of II to refer to sources that actually criticize them – cause concerns about bias.

For organizations that give priority to a politically neutral AI capable of providing unfiltered information, Grok 3 can be seen as inappropriate.

This dispute emphasizes the importance of the evaluation of both the technical capabilities and the main alignment of AI models before integrating them into business operations.

The search for truth is a victim of reputation management

The Grok 3 dispute assigned a wider debate over the development of AI, including whether AI models are aligned for the benefit of users or their creators and to what extent.

Critics say the internal is pushing, limiting criticism of Musk and Trump, pointing to a conflict of interest, especially given the ownership of Musk on X and XAI and Doge’s leadership.

In the meantime, AI’s ability to provide dangerous information emphasizes the ideologically and politically motivated nature of “equalization” when it comes to the Grok family of models.

In addition, he raises the question of how and how other AI models are prejudiced in favor of their creators or values ​​that are not shared by consumers.

At the same time, this gives consumers to pause when they are considering Grok 3 compared to the fast -paced market for alternative model AI models and models for reasoning such as OPENAI’s O3 series., R1 on Deepseek with open source., Gemini 2 Flash Thinking and others.