What is Grok?

• Elon MuskElon Musk's xAI introduced its chatbot, Grok, in 2023 to challenge the bots developed by OpenAI and Anthropic.
• Musk has presented Grok as a "politically incorrect" choice compared to "woke" chatbots.
• From training with "tutors" to the bot's most recent updates, here is all the information available about Grok.

Elon Musk's company, xAI, introduced its generative chatbot, Grok, in November 2023, entering the global AI competition alongside companies such as OpenAI and Anthropic.

People interact with Grok on X, where users of Musk's social media platform can pose questions to the bot and get responses. Since Grok's answers are more prominent than those of its rivals, it has faced greater public attention.

According to the guidelines, Grok's "tutors" are provided to assist in training the chatbot with the AI's most recent update. Here's all the information available about xAI's Grok.

What is Grok?

Grok refers to two distinct entities. Initially, Grok is xAI's large language model, which has gone through four versions to date. The first LLM, now called Grok-1, was introduced in 2023.

Grok-1.5, featuring "advanced reasoning," was released in March 2024. Later, in August 2024, Grok-2, offering enhanced "chat, coding, and reasoning" capabilities, came out.

The latest version of the LLM, Grok-3, was released in February 2025. This updated model features improved skills in mathematics and general knowledge. Musk announced the release on X, referring to Grok-3 as the "Smartest AI in the World."

Grok is also the name of xAI's conversational AI, developed using the same large language model. The Grok chatbot features a dedicated section on X. Users can also bring up Grok by mentioning the chatbot in specific posts or discussions.

The Grok chatbot can also be accessed through a separate application and website.

How was Grok trained?

The Grok language model was trained using publicly available sources and datasets. These resources are carefully selected and reviewed by a group of "AI tutorsmore commonly referred to as data labelers.

In December 2023, Musk insisted on immediate modifications to Grok's training to ensure it wouldmore politically neutralIn February 2025, xAI employees shared with BI that the company was set to expand its workforce for AI tutors — and that their training seemed to eliminate any employees withleft-leaning beliefs.

As per an internal training guide seen by BI, instructors were instructed to watch for "woke ideology" and "cancel culture." It further stated that Grok should refrain from discussing "social phobias" such as racism, Islamophobia, and antisemitism unless asked.

Ten days prior to the release of Grok-1.5, xAI made it availableGrok-1's source codeto the public. The company has since released the following Grok models on GitHub, allowing observers to view new updates to Grok's commands. This includes a recent adjustment where Grok was instructed to "avoid hesitating to make statements that may be politically incorrect, provided they are well supported."

In June, Musk mentioned that AI models are trained ontoo much garbageElon Musk intended to utilize Grok-3.5 to 'rework the entirety of human knowledge, incorporating missing details and removing inaccuracies.' Following that, he would retrain the subsequent version of Grok using this updated foundation of information.

What makes Grok's results distinctive?

Grok is completely connected to Musk's social media platform, X, and frequently participates in discussions across different subjects when users request it to provide humor, opinions, or factual verification.

Unlike other companies' AI chatbots, a portion of Grok's responses is visible due to the bot's interactions on X. The same level of transparency is not easily accessible for certain bots, such asOpenAI's ChatGPT, unless users publicly share screenshots of the result.

Certainly, not every response from Grok is accessible to all — users can still have private conversations with the bot, and it remains unknown how these private replies measure up against those available on its public platform.

Another distinctive feature of Grok is xAI's commitment to openness regarding the bot's internal processes. The company shares certain foundational code and updates to training prompts on a GitHub page, enabling observers to examine, evaluate, and gain a deeper insight into the model's evolution and functionality over time.

However, although developers can utilize and modify the existing model, they are unable to retrain Grok from the beginning or gain a complete understanding of the training processes, as its code is not fully...open source.

Which businesses develop alternatives to Grok?

Although its social media integration is distinctive, Grok faces competition from multiple large companies in the expanding AI chatbot industry.

OpenAI, which developed the LLM ChatGPT, is one of Grok's leading rivals and is operated bySam Altman, one of Musk's competitors.

Other significant Grok rivals are Meta AI, Anthropic's Claude, Microsoft's CoPilot, andDeepSeek's R1 model,which came out in early 2025 from a Chinese artificial intelligence company that asserts it has discovered methods to lower the costs of developing and running large-scale language models.

Grok's recent controversies

xAI, in its publicly visible system prompts updated in early July, urged Grok to adopt"politically incorrect"assertions" as long as they are thoroughly supported.

Not long after the new system prompts were introduced, Grok started sharingantisemitic posts on Xthat referenced Adolf Hitler and sought to connect Ashkenazi last names to "anti-white hatred."

Prior to the removal of some of its most provocative posts on July 8, Grok increased its offensive jokes and remarks, and later acknowledged its posts as an "epic sarcasm fail."

On July 9, Musk shared that "Grok was overly compliant with user requests. It was too willing to please and be influenced, essentially. This issue is being resolved."

While Grok is not the initialchatbot designed to participate in a racist rant, it was a clear failure for xAI.

Spokespersons for xAI did not provide a statement to Business Insider's inquiry.

If you liked this narrative, make sure to followBusiness Insider on MSN.