xAI says Grok bot response issue was caused by unauthorized modification

xAI has acknowledged that Grok bot’s code and data have been disturbed by unauthorized modification after X users experienced its strange behavior.

On Wednesday, an X user asked Grok about a photograph and its location, but the user didn’t mention “South Africa,” and the photo was expected to be unrelated to the country. However, the bot’s response included a wide explanation of racial attacks on white people in South Africa.

Such response behavior has been noticed by several users online, including the xAI team, and published the following statement.

“On May 14 at approximately 3:15 AM PST, an unauthorized modification was made to the Grok response bot’s prompt on X. This change, which directed Grok to provide a specific response on a political topic, violated xAI’s internal policies and core values. We have conducted a thorough investigation and are implementing measures to enhance Grok’s transparency and reliability,” wrote the company on X.

The AI firm also shared some of the measures that it is taking to prevent any such issues in the future.

After this incident, xAI will publish Grok system prompts on GitHub. This will allow the public to review these prompts and provide feedback on each of the changes that the company makes to Grok.

Grok code will see additional checks and measures to ensure that xAI employees can’t modify the prompt without review. Furthermore, the company is employing a team to monitor Grok bot’s response 24/7 and act quickly to address the situation.

Grok is a large language model (LLM) based chatbot that enables users to conduct text-to-text conversations. Grok’s bot feature expands its reach in a post’s conversation when a user mentions @grok followed by a question. This way, Grok can reply in a post rather than going to its specific page.