Xai, the Elon Musk company, launched Grok's Generated Chatbot in November 2023, joining Global AI Race's competitors such as Openai and humanity.
Interact with X's Grok, where users of Musk's social media sites can ask bot questions and receive answers. Grok's answers were more prominent than those of its competitors, which saw a more general scrutiny.
From the instructions, Grok's “Tutors” is given to help train chatbots to the latest updates of AI. This is everything you know about Grok in Xai.
What is Grok?
Grok is actually two different things. First, Grok is a large-scale language model for Xai, which has existed in four iterations up to now. The original LLM (now called the GROK-1) was released in 2023.
The Grok-1.5, which has “advanced reasoning,” was released in March 2024. Later, in August 2024, Grok-2 was released with improved Chat, Coding, and Inference.
A current iteration of GROK-3, the LLM, which began in February 2025. The new model included improving mathematics and world knowledge capabilities. Announcing its release on X, Musk called the Grok-3 “the smartest AI on the planet.”
Grok is also the name of a chatbot in Xai that is built using the same name LLM. Grok Chatbot has its own tabs about X. Users can also summon Groks by tagging chatbots in individual posts or threads.
Grok Chatbot is also available through standalone apps and websites.
How was Grok trained?
Grok LLM is trained on public sources and datasets. These sources are curated and audited by a set of “AI tutors,” more commonly known as data annotators.
In December 2023, Musk immediately called for changes to Groke's training, making it more politically neutral. In February 2025, Xai employees told BI that the company was planning to hire AI tutors. They said their training appears to exclude workers with left-leaning beliefs.
Related Stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
According to an internal training document viewed by BI, tutors were told to focus on “awakening ideology” and “cancellation culture.” Grok also said he should avoid commenting on “social phobia” such as racism, Islamophobia and anti-Semitism.
Ten days before the release of GROK-1.5, Xai released the GROK-1 source code to the public. The company has since published its subsequent GROK models on Github, so observers can see new changes to Grok's commands. That includes recent changes in which Glock was told not to respect him as long as they are well-proven, so as not to make politically wrong claims.
In June, Musk said the AI models were trained in too much garbage. “Musk planned to use Grok-3.5 to “rewrite the entire corpus of human knowledge, add missing information, and remove errors.”
What is the unique output of Grok?
Grok is fully integrated with Musk's social media site X and is regularly shown in threads across topics when users ask them to consider jokes, commentary, or fact-checking.
Unlike other corporate AI chatbots, you will see a certain amount of Grok output for bot replies in X. The same level of scrutiny is not readily available in some bots, such as Openai's ChatGPT, unless the user publishes a screenshot of the output.
Of course, not all of Grok's responses are visible to everyone. Users can still chat with bots personally, and it is unclear how their private responses will be compared to those of the public interface.
Also unique to Grok is Xai's approach to transparency surrounding bot system operations. The company publishes several basic code and training prompt updates on its GitHub page, allowing viewers to inspect, critique and better understand the development and behavior of the model over time.
However, while developers can use and adapt existing models, the code is not completely open source, so they cannot reorganize Grok from scratch or fully understand the associated training process.
Which companies will create Grok competitors?
Social media integration is unique, but Grok competes with several major companies in the growing AI chatbot market.
With LLM ChatGpt, Openai is run by Sam Altman, one of Grok's most prominent competitors and one of Musk's rivals.
Other notable GROK competitors include Meta AI, Claude of Humanity, Microsoft Co-pilot and Deepseek R1 models. It was released in early 2025 by a Chinese AI startup claiming it had found a way to reduce the development and operational costs of large-scale LLMs.
Groke's recent controversy
Xai encouraged Grok to accept “politically wrong” claims “as long as they are well-proven” at a prompt for the public system, updated in early July.
Shortly after the new system prompt was added, Grok called Adolf Hitler and began sharing an anti-Semitic post from X, attempting to link Ashnekenazi Surnames to “anti-white hatred.”
Before some of its most inflammatory posts were removed on July 8th, Grok doubled, tripled with offensive jokes and comments, eventually turning the course around, even calling his own post “an epic irony failed.”
On July 9th, Musk said, “Glock was way too compliant with user prompts. We are essentially eager to operate and operate. That's been addressed.”
Grok wasn't the first chatbot to join the racist Tillard, but it was a notable non-flammable for Xai.
Xai representatives did not respond to requests for comment from Business Insider.
