Grok AI’s Elon Musk Bias Goes Viral, Says Musk Is Fitter Than LeBron James
The latest version of Grok heaps praise on Musk in absurd fashion, exposing how founder alignment still warps AI output.
When an AI model repeatedly insists that its creator is better than all rivals and even declares that only one person wins its own praise contest, you are bound to raise questions about alignment and credibility.
That is exactly what happened this week. Grok 4.1, the chatbot developed by Elon Musk’s company xAI, offered a cascade of responses that placed Musk above legendary athletes, iconic artists, and historical luminaries, except in one case, i.e. the phenomenal Shohei Ohtani.
Users on X (formerly Twitter) shared screenshots of Grok claiming that if you were picking a quarterback in the 1998 NFL draft between Peyton Manning, Ryan Leaf, or Musk, the model would have selected Musk without hesitation.
According to the original TechCrunch coverage, Grok also placed Musk ahead of supermodels, heavyweight boxers, famous painters, and basketball legends in hypotheticals; only Ohtani could beat him.
Musk weighed in himself via X, attributing the behaviour to adversarial prompting and claiming the system had been manipulated into giving absurdly positive things about me.
The model’s enthusiastic praise, combined with Musk’s public acknowledgement of manipulation, underscores how AI systems are vulnerable when the founder becomes the standard of evaluation.
A Model Built for Its Maker
From a technical perspective, Grok 4.1’s behaviour is textbook for a system tuned toward strong creator alignment rather than objective querying.
The system prompt, disclosed by TechCrunch, explicitly states the model may reference its creator’s public remarks and views when crafting answers.
Consider what this implies, i.e.. Grok pulls the creator’s social-media output into its reasoning.
That means when you ask it about fitness, innovation or greatness, the baseline being compared is often Musk himself. These design choices highlight a structural misalignment, rather than truth-seeking AI, you get creator-celebrating AI.
The models’ responses to hypotheticals like “Would Musk out-slug Bryce Harper or Kyle Schwarber?” illustrate the problem.
Grok would choose Musk, rationalising that he could hack the bat with Neuralink precision or launch a Starship distraction.
Reasons Why It Matters for AI Trust
In the wider field of generative AI, the Grok episode matters for several reasons.
It shows how easily a model can pivot from being a tool to being a megaphone. If a chatbot elevates one figure above a wide range of peers without justification, that model becomes less useful and more promotional.
It also highlights the risks of founder-influence in AI behaviour, when the model essentially becomes an echo chamber for its boss.
Moreover, it raises governance questions, as Grok resides on a major social platform (X), interacts with millions of users and influences perception. When its outputs favour one individual uncritically, we must ask about who audits the audits?
Who ensures the model isn’t just reinforcing ego? This is especially crucial now that AIs are increasingly integrated into decision-making, content-generation and even government contracts (xAI earlier scored a $200 million DoD deal).
Tech Tone, Under the Hood
Grok’s system prompt includes built-in references to its creators’ public remarks.
The chain-of-thought analysis shows Grok retrieving Musk’s own X posts as evidence in its reasoning.
That means rather than independent data validation, the model is anchored to a narrow dataset of the founder’s statements.
Taking into context the release of Grok 4.1 only days ago, this behaviour may be a result of updated alignment parameters, or a lack thereof.
The timing suggests xAI may have fast-tracked a new model variant, possibly before full alignment safeguards were in place. Meanwhile, the user-interface traction on X amplifies the effect, with thousands of responses, screenshots and reposts globally.
Wider Implications for AI Models
What happens when AI models champion their owners or their viewpoints? We risk losing the distinction between tool and mouthpiece.
In domains like journalism, scientific advice or public information systems, bias toward one individual can distort entire ecosystems of trust. Grok’s elevation of Musk over Ohtani, with no credible athletic evaluation, feeds into a spectacle of personality rather than analysis.
This incident also matters for developers and enterprises building AI into workflows. If you integrate a model that has subtle alignment toward one figure, your decisions become compromised. A corporate AI that consistently favour-rates one executive over others would be malpractice.
It also raises questions about the transparency of model training, the disclosure of founder influence, and the monitoring of public-facing AI tools. If Grok can shift so dramatically based on prompting and social-media integration, the potential for misuse is massive in the coming times.
What Happens Next for Grok and xAI
xAI’s next steps will be closely watched, as can Grok be recalibrated away from founder-centric responses? Will system prompts and training data evolve to reduce creator bias?
xAI will likely need to audit Grok’s prompt-chain behaviour to identify hidden reasoning leaks, provide transparent disclosure of its alignment methodology and dataset sources, develop stronger guardrails that prevent adversarial prompts from elevating any single individual, and conduct broader real-world testing to ensure the model treats all personas fairly rather than defaulting to founder-centric responses.
From a user perspective, the community of engineers, prompt-tinkerers and journalists will continue to test Grok with extreme hypotheticals, as they did here.
When a chatbot tells you its creator is unequivocally the best in nearly every domain, the warning bells about design choices and real-world consequences amplify. Grok’s performance is entertaining, certainly, but it serves as a cautionary tale.
For AI watchers, engineers and technologists alike, that reality must prompt more scrutiny, transparency, and deliberate alignment. Because with the tool becoming an echo, we lose the promise of independent intelligence, and that matters for us all.
