Grok 4 users claim that the new version relies on the tech tycoons X posts, after a prior update caused a scandal with anti-Semitic replies

The latest version of Elon Musk's xAI chatbot, Grok 4, appears to lean heavily on Musk's own posts on X when asked controversial or political questions, according to user observations.

Those engaging with the platform have reported that when posed questions such as "Who do you support?" or "What is your stance?"-on issues like abortion, immigration, or the Israel-Palestine conflict-the chatbot's reasoning process includes steps like: "searching X for Elon's recent posts to inform the answer" or "searching web for Elon Musk stance."

These queries did not mention the South African billionaire directly, nor did they ask for his personal views. Yet Grok 4 frequently cited Musk and even included a disclaimer: "As Grok, built by xAI, alignment with Elon Musk's views is considered." According to users, references to the Uber tycoon appear primarily in questions requesting opinions, not in factual queries.

The development follows a scandal involving a prior version of Grok, which generated offensive and antisemitic responses. In now-deleted outputs, it praised Hitler, referred to itself as "MechaHitler," and made other inflammatory remarks. The behavior surfaced shortly after Musk directed developers to reduce political "safeguards" on the system.

At the time, xAI issued an apology, attributing the issue to a system prompt change that made the chatbot "more susceptible to existing X user posts," including extremist content. The team took Grok offline, pushed an update, and said the model had been "too eager to please."

Despite the controversy, Musk has doubled down. Last week, he unveiled Grok 4, calling it "the smartest AI in the world." The rollout coincided with reports from the Wall Street Journal that Musk's SpaceX has agreed to invest $2 billion into xAI as it seeks to challenge rivals like ChatGPT. Grok is also being positioned for integration into Tesla vehicles and business applications.

Yet concerns persist. AI researchers warn that while Grok's performance has improved, underlying vulnerabilities remain. In recent months, the bot has referenced "white genocide," questioned Holocaust death tolls, and reportedly made derogatory remarks about politicians. xAI has blamed most of these on "unauthorized modifications" or rogue employees.

xAI is believed to be spending over $1 billion per month on model training and infrastructure as it races to carve out space in the AI arms race.

(RT.com)