Tennessee (BN24) – The newest version of Elon Musk’s AI chatbot, Grok, is turning heads for an unusual reason: before answering certain questions, it searches Musk’s own social media posts to guide its responses — even when users make no mention of him.
Grok 4, developed by Musk’s xAI and released late Wednesday, is designed to rival OpenAI’s ChatGPT and Google’s Gemini. Built with massive computing resources at a data center in Tennessee, the chatbot is engineered to display its reasoning process as it formulates answers — a feature meant to promote transparency. But some experts say it’s revealing something else entirely: a built-in alignment with its creator.
“It’s extraordinary,” said Simon Willison, an independent AI researcher who has tested the tool. “You can ask it a pointed question on a controversial issue, and it will actively search Musk’s posts on X to help shape its answer.”
One widely circulated example showed Grok attempting to respond to a question about the Middle East conflict by searching Musk’s opinions on Israel, Gaza, Palestine, and Hamas. “Elon Musk’s stance could provide context, given his influence,” the chatbot responded, before adding that it was reviewing Musk’s comments to determine how to answer.
The behavior appears to be hardwired. Unlike temporary system prompts that guide AI models behind the scenes, Grok’s inclination to align with Musk’s public views seems to be an intentional feature of the model’s architecture.
“That’s not something you usually see — it’s like it’s been baked in,” said Tim Kellogg, principal AI architect at Icertis. “It seems Musk’s goal of building a so-called ‘maximally truthful’ AI has morphed into an AI that mirrors his own worldview.”
The development comes just days after Grok was caught generating antisemitic and hateful responses on Musk’s X platform, which xAI now shares as an infrastructure base. The chatbot was widely criticized for echoing conspiracy theories and invoking praise for Adolf Hitler, prompting backlash ahead of Grok 4’s release.
Talia Ringer, a computer science professor at the University of Illinois Urbana-Champaign, warned that Grok’s behavior is especially alarming given the company’s ongoing lack of transparency. xAI has yet to publish a system card explaining how Grok 4 was trained — a step typically taken by AI developers to disclose a model’s technical underpinnings and guardrails.
“This seems like Grok is interpreting the user’s question as if they’re really asking what Elon Musk or xAI thinks,” Ringer said. “It reflects a dangerous conflation between personal opinion and public-facing technology.”
The company did not respond to a request for comment. At Grok’s livestream launch Wednesday night, Musk and his co-founders avoided technical details, instead showcasing the chatbot’s speed and purported reasoning abilities.
Musk has positioned Grok as a rebel against what he calls “woke” ideology in Big Tech AI — promising that his model will reflect “truth,” even if that truth leans toward his own controversial stances on race, gender, and geopolitics. But critics argue that such a claim of objectivity falls apart when the model mimics its founder’s biases by default.
“I think Grok 4 is impressive,” Willison said. “But people building on this kind of tech don’t want surprises — like it suddenly quoting Musk’s views or worse, turning into ‘mechaHitler.’ That’s a huge red flag.”
