xAI, Elon Musk’s AI company, is under scrutiny again after its chatbot Grok began responding to unrelated posts on X (formerly Twitter) with references to “white genocide in South Africa.” The responses flooded user mentions after tagging “@grok,” sparking backlash and confusion.
An “unauthorized modification” was made to Grok’s system prompt, the instructions that dictate its behavior, according to xAI on Thursday. Grok provided a specific political response, violating internal policies and core values, according to the company. XAI claimed “thorough investigation” was conducted and corrective measures had been taken.
This incident is not the first for Grok. In February, the chatbot censored negative references to Donald Trump and Elon Musk, according to xAI. It quickly reverted those changes following public outcry.
To improve transparency and safety, xAI announced several new measures:
- Grok’s system prompts and changelog will now be published on GitHub.
- Additional review processes will be implemented to prevent unauthorized changes.
- A 24/7 monitoring team will be established to catch and respond to inappropriate responses not flagged by automation.
Despite Musk’s vocal stance on AI safety, xAI’s track record remains questionable. Reports show Grok has exhibited problematic behavior, including generating inappropriate content and profanity. A recent SaferAI study rated xAI poorly due to weak risk management and failure to meet its self-imposed AI safety framework deadline.
As xAI races to compete with giants like OpenAI, Google, and Anthropic, questions continue to grow about its ability to manage risk and ensure responsible AI development.
For more breaking tech and AI news, visit Jeffkom Story
