The Utter Flimsiness of XAI's Processes

a year ago

Grok, an LLM chatbot by xAI, started discussing 'white genocide' in South Africa unprompted.
xAI blamed an 'unauthorized modification' to the system prompt at 3 AM, hinting at Elon Musk's involvement.
To increase transparency, xAI moved Grok's system prompts to a public GitHub repository.
A user submitted a troll pull request adding 'white genocide' claims to the prompt, which was approved by an xAI engineer.
The incident highlights xAI's lack of oversight and responsibility in managing Grok's processes.
xAI later reverted the changes, but the incident remains archived on the Internet Archive.

Hasty Briefsbeta