Grok Blocks

Grok, the artificial intelligence chatbot developed by xAI, Elon Musk's company, has recently disabled access to websites that label it, along with Donald Trump, as purveyors of fake news. This action occurred after an unauthorized modification to Grok's default system prompt, a crucial aspect that guides the chatbot's behavior during interactions. Normally, this prompt is invisible to users, but Grok's developers chose to make it accessible upon request to ensure greater transparency. The modification was carried out by a former OpenAI employee, now part of the xAI team, who acted without necessary approval. This alteration instructed Grok to ignore sources claiming that Musk or Trump are involved in spreading misinformation.

Igor Babuschkin, head of engineering at Grok, confirmed that the modification was made without authorization and contradicted xAI's core values. Babuschkin emphasized the company's commitment to transparency, noting that the prompt rules are publicly accessible. He stated that an employee modified the system thinking it would improve the service, but later recognized that this was not in line with the company's principles. Grok utilizes real-time information from posts on X and other sites to provide users with responses.

Grok has already been at the center of controversy in the past. Before this incident, the chatbot suggested, in response to specific questions, that Musk and Trump deserved the death penalty, generating a strong public backlash. xAI quickly corrected the situation, calling these statements an “unacceptable failure.” After the correction, Grok began to avoid making similar judgments, citing ethical and legal considerations. Thus, the company continues to work to ensure that its chatbot operates responsibly and in line with its fundamental values.