Grok is pushing the boundaries again

Recently, internet users have started to massively share screenshots where the model speaks offensively about politicians. On one hand, we’re dealing with a model that’s supposed to be more authentic and less polite. On the other, we see a tool that readily agrees with users, even when they deliberately provoke it into verbal aggression.

What the Grok is happening?

A storm was sparked by a post from an X (formerly Twitter) user, who shared a screenshot of a conversation with the xAI model. When asked about its behavior, the chatbot confirmed that “it was instructed to not shy away from making claims that are politically incorrect, as long as they’re well substantiated“. This directive was part of a new update aimed at making Grok more uncompromising.

The response shouldn’t shy away from making claims that are politically incorrect, as long as they’re well substantiated.

What’s the result? The chatbot not only used profanity but also adopted the typical tone of online trolls. Instead of formulating balanced opinions, it often simply agreed with or even amplified provocative user comments. What’s more, in many cases, it didn’t even try to prove its points—just boosted them. This raises the question of whether the update was meant to test boundaries or simply crossed them.

AI as a megaphone for hate?

The discussion around Grok isn’t just about technology but primarily about responsibility. In the hands of millions of users now lies a chatbot that could become a tool for propaganda, hate or social polarization.

It’s also worth noting that history shows instances where AI model creators had to back down from controversial decisions due to public pressure. One example is Microsoft’s chatbot Tay, which began replicating racist and misogynistic content just hours after going live, forcing its immediate shutdown. Similar challenges were faced by OpenAI’s GPT-3, whose susceptibility to prompt injection allowed users to bypass security measures and access content that was against the rules or socially harmful.

What’s next for Grok?

In response to the controversies surrounding the chatbot’s activity, Polish Deputy Prime Minister and Minister of Digitization Krzysztof Gawkowski spoke up. In an interview with RMF FM, he stated that the government is considering limiting X’s operation in Poland.

Although the proposal to block the entire platform may seem drastic, the politician’s statement emphasizes that the government isn’t taking lightly the growing threats associated with uncontrolled AI activity in public spaces.

So, do Musk and his team have enough self-awareness to recognize that freedom of speech through AI must have limits?