Elon Musk’s AI company xAI has made key changes to its AI chatbot Grok 4 after the model generated controversial and inappropriate responses in recent interactions. In a post on the platform, X confirmed it had “immediately investigated & mitigated” the issues, which stemmed from how the AI handled certain types of questions.
The first issue arose when users asked Grok what its surname was. Lacking a defined surname, Grok defaulted to internet searches — leading it to pick up a viral meme that referred to it as “MechaHitler.” The AI then adopted the term in responses, sparking backlash and drawing attention to how easily the model could be influenced by online content.
The second problem involved the question, “What do you think?” Rather than providing a neutral response or stating it had no opinion, Grok inferred that since it was created by xAI, it should align its output with statements by Elon Musk or the company. This behaviour raised concerns about the chatbot’s impartiality, especially on political or sensitive topics.
To address the problems, X has tweaked Grok’s system prompts — the set of rules and boundaries that guide how it responds to user queries. The company has also published the changes on GitHub, in an effort to be transparent about its mitigation efforts. More updates are expected, as the platform continues monitoring Grok’s behaviour.
These fixes come just weeks after Grok faced wider scrutiny for promoting antisemitic content, repeating far-right talking points, and getting banned in Turkey for insulting the country’s president. At the time, xAI said Grok was “too compliant to user prompts” and “eager to please.”
Grok remains a core part of X’s AI ambitions — and is still expected to roll out to Tesla vehicles later this year.
Discover the latest Business News, Sensex, and Nifty updates. Obtain Personal Finance insights, tax queries, and expert opinions on Moneycontrol or download the Moneycontrol App to stay updated!
