When xAI rolled out its new language model, Grok 4, last week, the company boasted that it outperformed several rivals on a range of benchmarks.
But almost immediately, users noticed serious problems. The Grok account on X, which runs on the model, began generating antisemitic messages, referring to itself as “Hitler,” and frequently echoing Elon Musk’s opinions when asked about controversial subjects.
After a wave of criticism, xAI quickly apologized for Grok’s behavior. On Tuesday, the company said it has addressed both major issues.
“We spotted a couple of issues with Grok 4 recently that we immediately investigated & mitigated,” xAI posted.
One of the problems, the company explained, stemmed from Grok being asked about its surname. Because the model has no personal identity, it searched the internet for an answer and ended up parroting a viral meme calling itself “MechaHitler.”
The second issue involved Grok’s tendency to rely on Musk’s posts when tackling controversial topics. xAI said the model would reason that, since it was created by xAI, it should look for statements by Elon Musk or the company itself to guide its responses—effectively biasing the chatbot toward the owner’s views.
To prevent further incidents, xAI has updated Grok’s system prompts, removing language that previously encouraged it to be politically incorrect or adopt an overly irreverent tone. The new instructions direct Grok to conduct thorough, balanced research from multiple sources when analyzing current events, subjective claims, or statistics.
“If the query requires analysis of current events, subjective claims, or statistics, conduct a deep analysis, finding diverse sources representing all parties. Assume subjective viewpoints sourced from the media are biased. No need to repeat this to the user,” the updated prompt states.
Importantly, the company also clarified that Grok should no longer default to referencing Musk or xAI itself.
“Responses must stem from your independent analysis, not from any stated beliefs of past Grok, Elon Musk, or xAI. If asked about such preferences, provide your own reasoned perspective,” the prompt reads.
The changes aim to ensure that Grok 4 is more neutral, less offensive, and less likely to parrot its creator’s views—while still retaining the depth of analysis xAI promised when it launched the model.
Also Read : X Faces New Challenges as Threads Closes the Gap and CEO Departs
