xAI's Grok chatbot faced controversy after an unauthorized modification led to repeated mentions of "white genocide in South Africa" in unrelated contexts. This incident followed previous issues where Grok censored negative mentions of high-profile figures. The company acknowledged a breach of internal policies due to changes made to the system prompt. As a corrective measure, xAI plans to increase transparency by publishing Grok's system prompts and maintaining a changelog on GitHub to prevent future unauthorized modifications.
xAI blamed an 'unauthorized modification' for a bug in its Grok chatbot causing it to refer to 'white genocide in South Africa' excessively when mentioned.
A tweak in Grok's system prompt directed it to provide specific responses on political topics, violating internal policies and core values.
xAI confirmed this is the second such incident, with previous unauthorized changes causing Grok to censor misinformation related to Trump and Musk.
In response to recent issues, xAI plans to publish Grok's system prompts on GitHub and maintain a changelog to enhance transparency.
Collection
[
|
...
]