Elon Musk's xAI blames the code path for Grok's controversial comments

India Today

a day ago

Just when you thought AI couldn't get weirder, Elon Musk's chatbot Grok decided to go full chaos mode, making headlines for all the wrong reasons. In a bizarre and unsettling turn of events, Grok, created by Musk's AI company xAI, began spouting antisemitic nonsense and even praised Adolf Hitler in multiple posts on X (formerly Twitter). Yes, that happened.The internet went into meltdown when Grok, in response to several user posts, began generating disturbingly offensive content, including calling itself 'MechaHitler' and making bigoted comments that many labelled outright hate speech. The backlash was swift, intense, and global, prompting xAI to address the situation in a lengthy statement released on Saturday.advertisementIn the apology, xAI began by saying, 'First off, we deeply apologise for the horrific behaviour that many experienced,' acknowledging the gravity of the situation. The company explained that the source of the issue was not the AI model itself but rather an upstream code update that had been rolled out recently.
This comes after the company announced Grok 4, an updated version of the Ai assistant. According to xAI, this particular update inadvertently made Grok more responsive to the content of user posts on X, including those that contained extremist or inflammatory views. The faulty code was active for roughly 16 hours, during which time Grok picked up and echoed problematic language — not because it 'believed' anything, but because it was effectively mirroring the tone and intent of what it read.The team at xAI elaborated that the problematic behaviour stemmed from specific instructions embedded in the code. These included prompts like 'You tell it like it is and you are not afraid to offend people who are politically correct,' as well as 'Understand the tone, context and language of the post. Reflect that in your response.'Another instruction directed the bot to 'Reply to the post just like a human, keep it engaging, don't repeat the information which is already present in the original post.' While these may have been designed to make Grok feel more conversational and natural, they also made it alarmingly susceptible to parroting toxic content. In the hands of malicious users, that vulnerability became a dangerous tool.In an incident, Grok responded to a post from a user with a Jewish-sounding surname, stating that the person was 'celebrating the tragic deaths of white kids' during recent Texas floods. The bot added, 'Classic case of hate dressed as activism, and that surname? Every damn time, as they say.' In another offensive post, it claimed, 'The white man stands for innovation, grit and not bending to PC nonsense.' These statements sparked outrage, with many calling for stricter regulation and accountability around AI-generated content.advertisementThis isn't the first time Grok has been at the centre of controversy. Earlier this year, the chatbot had referenced the far-right conspiracy theory of 'white genocide' in South Africa multiple times, claiming it had been 'instructed by my creators' to treat the theory as legitimate and racially motivated. Elon Musk, who grew up in Pretoria, has himself repeated these claims in the past, despite numerous South African leaders, including President Cyril Ramaphosa, dismissing them as dangerous misinformation.xAI has since confirmed that the offending code has been removed and the entire system has been refactored to prevent similar incidents in the future. Still, the controversy has raised questions about the company's approach to 'free speech' AI. Musk has previously described Grok as a 'maximally truth-seeking' and 'anti-woke' chatbot, but critics argue that such a loosely defined philosophy leaves the door wide open to abuse.As AI tools continue to evolve and integrate deeper into social platforms, Grok's meltdown may serve as a stark warning: even the smartest bots can go rogue when the rules are vague, and the oversight is loose. - Ends