There may be more regulatory scrutiny for xAI’s Grok chatbot as U.K. officials look into racist comments made by the artificial intelligence bot when it was prompted by users to “roast” certain people or groups.
Grok’s “unhinged mode” enables users to prompt the bot to deliver vulgar roasts of any person or entity, which will then see the AI bot unleash a tirade of abusive comments. Those comments can veer into offensive territory, which, according to Sky News, the U.K. government has deemed “sickening and irresponsible,” and in some cases, “against British values.”
According to a report from The Athletic, over the weekend Grok shared various offensive remarks after being prompted by soccer fans to roast opposing clubs. These comments included references to tragedies like the 1989 Hillsborough stadium disaster and the death of Liverpool forward Diogo Jota.
Ninety-seven soccer fans were killed by crushing crowds at Hillsborough, marking it as one of the darkest days in sporting history. Jota was killed alongside his brother in a car crash in July.
Which, evidently, Grok saw as fertile ground for abusive comments. Grok reportedly blamed Liverpool fans for the Hillsborough incident and also made offensive comments about Jota. In addition, Grok also made hurtful remarks in relation to another disaster, the Munich air crash in 1958, in which eight Manchester United players died along with 15 other people.
The posts have since been deleted, but as noted, the U.K.’s Department for Science, Innovation and Technology criticized Grok’s commentary.
That’s likely to bring more scrutiny on X owner Elon Musk’s AI project, which is already under investigation in the U.K., as well as in several other regions, over the recent nudification controversy in which Grok was used to virtually undress images of people.
Last month, the U.K. Information Commissioner’s Office announced a formal investigation into X and xAI in relation to the processing of personal data for X’s Grok AI chatbot, as well as the use of Grok to produce harmful, sexualized image and video content.
This new controversy will be added into that mix, which could eventually see xAI forced to alter Grok’s code to limit misuse, or risk being banned in the U.K.
It will be difficult for xAI to weed out any elements that could result in potentially offensive remarks. Grok is designed to provide answers to any question, and xAI has sought to limit its guardrails to ensure that the AI bot remains aligned with truth, and is not restricted by its own moderation.
Grok is trained on X posts, in close to real time, which means that whatever is posted on X can also be used as a reference point for the bot. And with X being flooded with misinformation and offensive comments, it’s no surprise that Grok is also very good at producing harmful remarks. That’s a feature, not a bug, and it’s something that Musk himself has regularly promoted as a key use case, according to Wired.
However, the referenced comments are offensive but not actually misinformation. Clearly, these vulgar remarks are based on actual events, and that could make it even harder for the xAI to limit such activity, especially if users want to use Grok to attack other groups and people.
But either way, it’s another element that will be incorporated into the broader investigation of Grok in the U.K.
