Elon Musk’s xAI acknowledges security breach in Grok bot: What happened and what’s next

HIGHLIGHTS

Elon Musk’s AI company, xAI, has confirmed a security breach involving its chatbot Grok.

The incident, which happened on May 14, reportedly caused Grok to repeatedly post responses about 'white genocide in South Africa.'

"An unauthorised modification was made to the Grok response bot's prompt on X," xAI said.

Elon Musk’s xAI acknowledges security breach in Grok bot: What happened and what’s next

Elon Musk’s AI company, xAI, has confirmed a security breach involving its chatbot Grok. The incident, which happened on May 14, reportedly caused Grok to repeatedly post responses about ‘white genocide in South Africa,’”’ even in response to unrelated posts on X (formerly Twitter).

In a post shared this morning on X, xAI explained what caused the issue. “On May 14 at approximately 3:15 AM PST, an unauthorised modification was made to the Grok response bot’s prompt on X,” the company wrote. “This change, which directed Grok to provide a specific response on a political topic, violated xAI’s internal policies and core values,” it added.

xAI said it has conducted an investigation into the matter and is now taking steps to improve transparency and prevent similar incidents in the future.

Also read: Snapdragon 7 Gen 4 chipset for mid-range smartphones launched: Here’s what it offers

Here’s what xAI is doing next:

  • Publishing prompts publicly: xAI will now make Grok’s system prompts available on GitHub. The public will be able to review those prompts and provide their feedback. “We hope this can help strengthen your trust in Grok as a truth-seeking AI,” the company said.
  • Better review process: The company admitted that its regular review process for changes was “circumvented” in this case. To stop this from happening again, xAI will add extra checks so that no one at the company can make changes to the bot’s prompt without review.
  • 24/7 monitoring team: xAI is also setting up a 24/7 monitoring team to respond to unusual or incorrect responses from Grok.

Also read: OpenAI unveils GPT-4.1 series with faster coding and better instruction following

The incident highlights how even advanced AI systems can go off track if internal controls are not properly followed.

Also read: Used Siri between 2014 and 2024? You could get up to Rs 8,500 in Apple Siri settlement case, here’s how

Ayushi Jain

Ayushi Jain

Tech news writer by day, BGMI player by night. Combining my passion for tech and gaming to bring you the latest in both worlds. View Full Profile

Digit.in
Logo
Digit.in
Logo