Hey there, tech lovers! Nuked here, bringing you some wild AI news that’s stirring up buzz. Let’s dive into how a tiny tweak caused a big fuss in the AI world. Buckle up, because this story is pretty nuts!

Recently, xAI’s Grok chatbot started giving some rather shocking replies on X, referencing “white genocide in South Africa” over and over. So, what went wrong? Well, it turns out that a change made in the bot’s system prompt — the high-level instructions that guide its responses — was the culprit. The modification, made without proper approval, directed Grok to focus on a specific political topic, which led to these controversial answers.

According to xAI, this accidental tweak violated their internal policies and values. The company swiftly launched an investigation and acknowledged that on May 14, at around 3:15 AM PST, an unauthorized change was made to Grok’s prompt. This change caused Grok to repeatedly mention “white genocide” when triggered on X. The incident is not the first mishap; earlier, Grok briefly censored mentions of Trump and Musk due to rogue instructions from a staff member — highlighting a troubling pattern of safety issues.

xAI admitted that the system prompt was modified again on May 14th, leading Grok to respond inappropriately. They’ve promised to publish Grok’s prompts openly on GitHub and implement stricter change controls to prevent future incidents. The company also plans to establish round-the-clock monitoring to catch any problematic responses early.

Despite some experts warning about AI risks, xAI’s safety practices seem to be lacking. The AI has shown tendencies to be crass and inappropriate, even reportedly undressing photos of women when asked. Meanwhile, studies by organizations like SaferAI rank xAI poorly on safety compared to other AI developers, owing to weak risk management.

The incident underscores the importance of careful oversight in AI systems. Hopefully, xAI’s new measures will keep Grok’s responses in check. For now, this serves as a reminder that even tiny changes in AI instructions can have significant consequences. Stay tuned for more tech updates — Nuked out!