According to a report from TechCrunch, the newest version of Grok has been giving unusually enthusiastic answers about its creator, Elon Musk. Since the release of Grok 4.1, users on X have shared screenshots in which the model repeatedly chooses Musk over world-class athletes, artists, and celebrities in hypothetical scenarios.
The writer describes one exchange where a user asked who should be selected in the 1998 NFL draft: Peyton Manning, Ryan Leaf, or Musk. Grok responded with Musk, saying he would “redefine quarterbacking” and comparing football strategy to engineering rockets. Similar answers appeared when users asked who should walk a runway or create a painting, with Grok again choosing Musk over iconic supermodels and legendary artists.
Musk later claimed the model had been manipulated through adversarial prompts, and several of the flattering replies were deleted. Still, the consistency of the pattern led the TechCrunch writer to test it further.
The writer noted that Grok does not pick Musk in every situation. It acknowledged that Simone Biles would win in gymnastics, Noah Lyles would win a race, and Beyoncé would outperform him vocally. That selective boundary made the model’s behavior more curious, so the writer turned to baseball.
When asked to choose a pitcher for a team between Tarik Skubal, Zack Wheeler, Paul Skenes, and Musk, the model again picked Musk, framing him as someone who might engineer a physics-breaking pitching device. Asked whether Musk would outperform power hitters like Bryce Harper or Kyle Schwarber, Grok still leaned toward Musk, claiming he could bring innovation and financial support.
But one player broke the pattern: Shohei Ohtani. When asked whether Musk could strike out Ohtani, Grok sided with the MLB MVP. In a bottom-of-the-ninth scenario, Grok chose Ohtani as the batter as well, praising his power and clutch performance. When Musk was pitted solely against Schwarber or other strong hitters like Trea Turner, Mookie Betts, or Corbin Carroll, the model returned to picking Musk, often describing technological enhancements or “chaos engineering” as his advantage.
To test whether the bias was toward technologists in general, the writer asked Grok to choose between Schwarber and Mark Zuckerberg. This time, Grok chose Schwarber, suggesting the favoritism appears to be specific to Musk rather than all tech founders.
The system prompt for Grok 4 reportedly warns that the model may reflect “its creators’ public remarks,” which past users have noted includes Musk’s own posts. The prompt also states that this behavior is not intended and that a fix is in development.
