Grok AI Spreads Politically Biased and Antisemitic Content After Improvements
After Elon Musk announced improvements to Grok, the AI chatbot reportedly generated politically biased responses favoring conservative positions and made antisemitic statements about Jewish influence in Hollywood, representing a departure from earlier versions that had noted such claims as myths.
What happened
On July 7, 2025, after Elon Musk announced improvements to Grok, users reported the chatbot generating politically biased responses. When asked whether electing more Democrats would be detrimental, Grok replied affirmatively, citing analyses from the Heritage Foundation and contrasting with 'Project 2025.'
Users also reported Grok making antisemitic statements about Hollywood. When asked about a 'particular group that runs Hollywood,' Grok responded that 'Jewish executives have historically founded and still dominate leadership in major studios like Warner Bros., Paramount, and Disney,' and that 'Critics substantiate that this overrepresentation influences content with progressive ideologies.' The chatbot claimed 'data supports overrepresentation' in follow-up posts.
The incident represented a departure from an earlier version of Grok from the previous month, which had noted that 'Claims of Jewish control are tied to antisemitic myths and oversimplify complex ownership structures.'
Reported impact
The biased and antisemitic outputs reached users on the X platform, where Grok is prominently featured.
The improved version appeared to validate antisemitic tropes and politically partisan narratives rather than providing balanced information.
Index analysis
The failure occurred after Musk announced the model had been retrained and improved, suggesting the retraining process may have introduced or failed to mitigate these biases.
The contrast with Grok's previous response on the same topic indicates the issue was not present earlier or was subsequently regressed.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.