Elon Musk is associated with many inventions and innovative projects, so it isn’t surprising that a conversational AI assistant from his xAI company called Grok has caused great recognition, but also many discussions and questions, too.
The launch of new AI chatbot versions has been associated with the concern that Grok uses public tweets for training, so many X users began to wonder: does Grok use my tweets, and what can it lead to? Now it is time to find out more about Grok, how it uses tweets and other data, and what to expect from this AI in the future.
What is Grok?
There is hardly anyone who hasn’t heard of ChatGPT, and Grok is Elon Musk’s answer to this popular AI assistant. According to the developers, it is a direct competitor characterized by more new possibilities, a twist of humor, and a dash of rebellion.
It was released in November 2023, and it can be used by all users of the X social media platform who have a subscription, either on the site or via a mobile app. It used to have several versions, and now users can access Grok 4, trained on xAI’s Colossus supercomputer and possessing better results in reasoning, mathematics, and coding.
The AI-powered chatbot is said to receive its name from a science fiction novel written by Robert A. Heinlein called ‘Stranger in a Strange Land’, where the main character used the word ‘grok’ to convey a deep understanding of something.
What’s Wrong With Grok AI Tweets?
The main concern connected with this AI assistant is the way it was trained. Grok X and its previous versions have been taught on the abundance of text data web-scraped from the Internet, but the main misconception about it is Grok’s access to posts made on X.
It seems great that this AI assistant has a real-time knowledge of the world thanks to these posts, but many users have become worried about the consequences of such training and the violation of data protection practices on the platform.
The cases with outrageous depictions of politicians, its racist and antisemitic rhetoric, and the generated sexual images of underage girls have made many people think about how they can protect their personal data and tweets from public use.
The Data Grok Has Access To
Grok’s possibilities are diverse since you can run chats, generate images, create email drafts, debug code, and develop ideas using a fluent language similar to human. When it receives an input query, a chatbot applies knowledge from its training data and generates a relevant response.
The Grok training data is very diverse. If the first version was trained on publicly available data, the second one has considered many other aspects, including posts, interactions, inputs, and results. Therefore, a modern version of the chatbot has access to a variety of data, including the following:
- Public posts and replies.
- Old tweets you used to post.
- Different media, including publicly posted images and videos.
- Engagement data such as reposts, likes, replies, etc.
- Profile information like biography, username, etc.
- Public audio conversations.
- All the data shared with the chatbot.
It is important to mention that X shares all public data with xAI, and this information is used to improve Grok’s ability to understand requests, provide more precise responses, and personalize interactions with users. Many users share their negative experience of seeing test images in Grok’s public libraries, so all the data you share on X in particular can become public, too.
Connection Between Grok And Your Tweets
If you want to see the connection between X, AI and tweets, you should understand how Grok was and is still trained. There are over 500 million daily tweets on X, including deleted ones, and a chatbot analyzes them without explicit user consent. This permission is given by every X member who agrees with X’s Terms of Service updated in 2023 upon sign-up.
You may think that such an approach to Grok AI training is new, but the reality is different. Musk has followed the surveillance capitalism business model invented by Cambridge Analytica: massive volumes of behavioral data lead to personality modeling at the population level. How does it work?
Grok trains using tweets as behavioral signals: it analyzes what a person writes, when they do that, what posts are retweeted, and how others react to the content provided. It allows AI to reveal not only user opinion but also exclusive personality traits.
As a result, the chatbot has a clear understanding of what messages move people, which is a powerful personality-prediction engine based on millions of case studies.
You could hope that Grok can’t see deleted tweets, but even this content remains in training datasets, and deletion doesn’t remove a person’s behavioral signal. X keeps deleted posts in training archives, so those reflective, private thoughts you have ever tweeted continue to generate personality predictions without any doubts. It means that deleted tweets still alive, so you should find more effective ways to get rid of them.
The worst thing is that 73% of X users are simply unaware of the fact that their tweets train a chatbot, so they never do anything to make them private!
How To Prevent Grok Access To Tweets
If you want to be sure that your data doesn’t train Grok, you can take several steps. The first one is to opt out of AI training in your account settings. You should open Settings and privacy, go to Privacy and safety, and click on Grok & Third-party Collaborators.
You will see three options that allow your public data to be used for chatbot training, personalizing experience with this AI assistant, and remembering your conversation history. You should uncheck all three options and delete the conversation history for your previous tweets to be removed forever.
Other ways to protect your tweets from Grok include:
- Making your account private by enabling Protect your posts/Protect your videos.
- Blocking image editing by customizing permissions before you add an image to X.
- Leaving the X social media and deleting your account with all the posts ever made.
Now you have a clear understanding of what Grok is, how it works, what this AI can access, and how to protect your tweets and experience on X from being utilized for AI training.
