Last April, billionaire Elon Musk revealed that his company X's AI chatbot Grok would introduce a mysterious "unrestricted mode." Nearly a year later, xAI, the company overseeing Grok's operations, updated its FAQ page on the website, providing details about this new mode.
In unrestricted mode, Grok is designed to deliver responses that are "intended to be offensive, inappropriate, and controversial," akin to "an amateur stand-up comedian still honing their craft," according to the page's description.
At present, this mode does not appear to be live yet.
The unrestricted mode may represent Musk's attempt to fulfill Grok's original vision. Around two years ago, when Musk announced Grok, he positioned it as avant-garde, unfiltered, and anti-"political correctness," willing to tackle controversial questions other AI systems avoid. To some extent, Grok has delivered on this promise. For instance, when asked to use vulgar language, Grok complies readily, using coarse and vivid language that chatbots like ChatGPT wouldn't employ.
However, the current version of Grok remains somewhat reserved when it comes to political topics, avoiding certain boundaries. In fact, a study found that Grok leans slightly left on issues such as transgender rights, diversity programs, and inequality.
Musk attributes this behavior to Grok's training data—public web pages—and has pledged to "make Grok more politically neutral."
In a December 2023 post, Musk wrote, "Unfortunately, the internet, which serves as Grok's training data, is rife with politically correct nonsense. Grok will improve. This is just the beta version."
